{
  "artifactKind": "gemma4_31b_af16_doppler_selected_logit_splice_receipt",
  "blockers": [],
  "claim": {
    "notWhat": "Not a full-vocabulary argmax, not a full layer-59 CSL run, and not hardware execution. It avoids the full-logits copyback wall by binding the Doppler top-k candidate logits; strict logit tolerance status is recorded separately from the top-token decision bound.",
    "scope": "Doppler supplies real Gemma 4 31B af16 post-FFN state for the final prompt position; CSL computes final RMSNorm plus the tied lm-head logits for the Doppler top-k token candidates across hidden chunks and preserves the top-token decision."
  },
  "comparisonMode": "argmax_decision_bound",
  "cslRun": {
    "allCandidateLogitsWithinTolerance": false,
    "argmaxDecisionStable": true,
    "atol": 0.02,
    "candidateCount": 32,
    "candidateRuns": [
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 13.98351001739502,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy",
              "sha256": "340351e8d251f0d633227deaa299ac93c0ea5cdbc05260087b89471a7f44c63a"
            },
            "outputSha256": "340351e8d251f0d633227deaa299ac93c0ea5cdbc05260087b89471a7f44c63a",
            "outputValue": 13.983509063720703,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/weight.npy",
              "sha256": "2d23d186a027a06ae5b3ca5076844bf22d646145b6c66a65ecb55ed855f8a0bd"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 15.017446517944336,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy",
              "sha256": "83530e337324211aad835854eddc482634f71200ab6a1bfdaf0e32c40abac15d"
            },
            "outputSha256": "83530e337324211aad835854eddc482634f71200ab6a1bfdaf0e32c40abac15d",
            "outputValue": 15.017444610595703,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/weight.npy",
              "sha256": "5623016a9e6cc8774e996dba7e617c6c29421f283f2b501c2926a125a59ad2c7"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.004585266113281,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy",
              "sha256": "3829aaf82a1ff1e1727a03186da677c943c4539935b52532457d6161b6f6b5fc"
            },
            "outputSha256": "3829aaf82a1ff1e1727a03186da677c943c4539935b52532457d6161b6f6b5fc",
            "outputValue": 7.004586219787598,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/weight.npy",
              "sha256": "9682b6c538b0ab1eeefdc084c15ea71e72ae06903c9c9e573012b0db1a738011"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 13.789728164672852,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy",
              "sha256": "f8375119663126093a1790b0d1eeb72bf8a1c35ff7080898f182660001593b6e"
            },
            "outputSha256": "f8375119663126093a1790b0d1eeb72bf8a1c35ff7080898f182660001593b6e",
            "outputValue": 13.789728164672852,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/weight.npy",
              "sha256": "695a88ac3ee08948f1f00de41984183087523f79330221311f241408b51fa8eb"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.878942489624023,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy",
              "sha256": "06428400630673a00f1fc6316780d30e56e93773761d338988881cf203688fd9"
            },
            "outputSha256": "06428400630673a00f1fc6316780d30e56e93773761d338988881cf203688fd9",
            "outputValue": 4.87894344329834,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/weight.npy",
              "sha256": "daa29d2e6e0fca7c94f9169594725b04d9f5580160c83e6ff28930ccc44dca98"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 2.9069056510925293,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy",
              "sha256": "a09ee00c22b4474bbfb9045c987eb39976f2a4b2df5d06e18412a683d11df99c"
            },
            "outputSha256": "a09ee00c22b4474bbfb9045c987eb39976f2a4b2df5d06e18412a683d11df99c",
            "outputValue": 2.9069058895111084,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/weight.npy",
              "sha256": "bf8782b5ebda2d0b0f7c6154467fa4d27bb6b9de22d41fb67e448f0ecf0a651c"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 57.58111572265625,
        "cpuSoftcappedLogit": 28.735962280829014,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 28.744709014892578,
        "lmHeadRowSha256": "2ebf2c44877f6ad0ca24484b636db369aca76bea856ae383b8dde725fb707032",
        "logitAbsDiff": 0.008746734063564077,
        "rawLogit": 57.58111572265625,
        "referenceRank": 1,
        "softcappedLogit": 28.735962280829014,
        "tokenId": 3730
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.405414581298828,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/output.npy",
              "sha256": "4cc768739d26701bd2019c87aafee97616d75c8b15411d6766a0a18b918d1d2d"
            },
            "outputSha256": "4cc768739d26701bd2019c87aafee97616d75c8b15411d6766a0a18b918d1d2d",
            "outputValue": 5.405414581298828,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0000/weight.npy",
              "sha256": "a71bea4f5475e7e9679f1219f3a46e780274ea83be0f4270ac925344b3eec72d"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.303387641906738,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/output.npy",
              "sha256": "94a3b89a62b02c13a229a855438a45264e7b67a79afde6ebd8a668b04b647270"
            },
            "outputSha256": "94a3b89a62b02c13a229a855438a45264e7b67a79afde6ebd8a668b04b647270",
            "outputValue": 8.303387641906738,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0001/weight.npy",
              "sha256": "4769525cad513fe72eb4b5f25f71d1fbedf3e35200aeee2b06b57b3ce2f2a044"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.431512832641602,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/output.npy",
              "sha256": "7f6468ecf8abbbb2a1bd186881f770bda85b164095840123a41dd8b455161ea1"
            },
            "outputSha256": "7f6468ecf8abbbb2a1bd186881f770bda85b164095840123a41dd8b455161ea1",
            "outputValue": 4.431513786315918,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0002/weight.npy",
              "sha256": "a783eb9b441e839ed8c0599056cfbef66604938e0b73074b7b7db5f59111e4ae"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.651602268218994,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/output.npy",
              "sha256": "a6316f9c178ae89dea5dcd7bed9434a2120b8345d87c6f8a250eb1a49355b4b9"
            },
            "outputSha256": "a6316f9c178ae89dea5dcd7bed9434a2120b8345d87c6f8a250eb1a49355b4b9",
            "outputValue": 7.651602268218994,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0003/weight.npy",
              "sha256": "db067e7cb6c1aa2dc40379a8acb661fa528d9c3246a01b91d1ab7c3563aec103"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.269284248352051,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/output.npy",
              "sha256": "2eb2936e9ba0600eb9a648de8e37e8d58bd3c3dba5273bb00685f135e58316da"
            },
            "outputSha256": "2eb2936e9ba0600eb9a648de8e37e8d58bd3c3dba5273bb00685f135e58316da",
            "outputValue": 7.269284248352051,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0004/weight.npy",
              "sha256": "f87faef401be7c89befbb1a5abb9933783f0e6889e3b58743d18ff8ccd11bbbe"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 2.2941155433654785,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/output.npy",
              "sha256": "3178a5bf0eb68d2384ad637187c6c6d2912c8b7349817316fd66547659b321fc"
            },
            "outputSha256": "3178a5bf0eb68d2384ad637187c6c6d2912c8b7349817316fd66547659b321fc",
            "outputValue": 2.2941155433654785,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-4781/chunk-0005/weight.npy",
              "sha256": "27f3ddcefaff2d38cf26db95975e721263ad6773011886ade783681b0de4ad40"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 35.355316162109375,
        "cpuSoftcappedLogit": 24.809441075123132,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 24.83536148071289,
        "lmHeadRowSha256": "03c687533b180ada318278d472820d56419293d1afc868cfefcc7ae2b879a25a",
        "logitAbsDiff": 0.025919199757154843,
        "rawLogit": 35.35531997680664,
        "referenceRank": 2,
        "softcappedLogit": 24.809442280955736,
        "tokenId": 4781
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.0041656494140625,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/output.npy",
              "sha256": "0d1c244021d5d5add79414f6fddf500c26ad2c887386a97d201039083dfb77a7"
            },
            "outputSha256": "0d1c244021d5d5add79414f6fddf500c26ad2c887386a97d201039083dfb77a7",
            "outputValue": 7.0041656494140625,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0000/weight.npy",
              "sha256": "b6463c69e293b0729049af829de05ca1fbe73e88e4e61566012bec02742c0454"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.6799421310424805,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/output.npy",
              "sha256": "f17332a0d9b03eff9ea480f4899f36b296ec610b9d671ab54976074ba517819e"
            },
            "outputSha256": "f17332a0d9b03eff9ea480f4899f36b296ec610b9d671ab54976074ba517819e",
            "outputValue": 6.6799421310424805,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0001/weight.npy",
              "sha256": "dc4f4f261d4baca61359579737736c0566738c9b741e40c088fca83cd9487313"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.621387481689453,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/output.npy",
              "sha256": "7b0bb7700a0cea7cacfa5efcb869ff0579a34c1139faeade974b47614fe0dbc3"
            },
            "outputSha256": "7b0bb7700a0cea7cacfa5efcb869ff0579a34c1139faeade974b47614fe0dbc3",
            "outputValue": 8.621386528015137,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0002/weight.npy",
              "sha256": "6004f766120fdf76f3d3e968126ae9ba252a0e239d0ebcde3e71d2affa22d612"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 13.064391136169434,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/output.npy",
              "sha256": "4e4a3993597709f9e449a4d0bdd8e361c0a691aa151a1c376e65e75bbdc35800"
            },
            "outputSha256": "4e4a3993597709f9e449a4d0bdd8e361c0a691aa151a1c376e65e75bbdc35800",
            "outputValue": 13.0643892288208,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0003/weight.npy",
              "sha256": "4afda96bc65af6d587c6d557d4e4486db4cd63dd407292aaf8c7f5b306247b5d"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -2.76796817779541,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/output.npy",
              "sha256": "b65f8fd8cafbdb9377f5d2d65e412eefe6ad34ae3d46f6dfb314c9007293486b"
            },
            "outputSha256": "b65f8fd8cafbdb9377f5d2d65e412eefe6ad34ae3d46f6dfb314c9007293486b",
            "outputValue": -2.76796817779541,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0004/weight.npy",
              "sha256": "0106a245c5e992e981c40ba7ca8f6f079cd7416f425f2dcf20cb01ae74e52eaf"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.9039696455001831,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/output.npy",
              "sha256": "655ad8323d55ddf72dc3ab601f1f7d281d6c31f553de3a5ce4f23d3648bd0999"
            },
            "outputSha256": "655ad8323d55ddf72dc3ab601f1f7d281d6c31f553de3a5ce4f23d3648bd0999",
            "outputValue": 0.9039695858955383,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5213/chunk-0005/weight.npy",
              "sha256": "6945a0df869b312306945667e3b72e814a743e01c97727a07b1c9e4186340514"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 33.505889892578125,
        "cpuSoftcappedLogit": 24.1942499179445,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 24.18125343322754,
        "lmHeadRowSha256": "f82dc96af0c0da89f87ae82bb94088b338839cb9f025b64b1d4b960b5168fc04",
        "logitAbsDiff": 0.012995151105990033,
        "rawLogit": 33.50588607788086,
        "referenceRank": 3,
        "softcappedLogit": 24.19424858433353,
        "tokenId": 5213
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.7877655029296875,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/output.npy",
              "sha256": "735ed3bb4b630d75440fa73cf57a6c941948cb3aef41a36a4ffc8d2ce8c53f94"
            },
            "outputSha256": "735ed3bb4b630d75440fa73cf57a6c941948cb3aef41a36a4ffc8d2ce8c53f94",
            "outputValue": 6.787764549255371,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0000/weight.npy",
              "sha256": "8be39b8f381a89ed860af215e0b8900fdefc1fc6cb3ffa76a5fdbe58af067a95"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.832139015197754,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/output.npy",
              "sha256": "af71516dfb057817d7b1a97d00eb330394250e70e3aeace7bfcfb2306802ca0b"
            },
            "outputSha256": "af71516dfb057817d7b1a97d00eb330394250e70e3aeace7bfcfb2306802ca0b",
            "outputValue": 5.8321380615234375,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0001/weight.npy",
              "sha256": "16e2df65bed7d92f9a174cc534ae545aa800d60574b623cb3795d2483be58792"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.477313041687012,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/output.npy",
              "sha256": "9b152397b27b15cfa11c5ce5c29ede744a1f4e27a8ad96f86944cbad92903479"
            },
            "outputSha256": "9b152397b27b15cfa11c5ce5c29ede744a1f4e27a8ad96f86944cbad92903479",
            "outputValue": 8.477312088012695,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0002/weight.npy",
              "sha256": "45786d63bc67085aaa1de95bc22232a639f5cf177442dcaac9478da891349dbc"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.017836570739746,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/output.npy",
              "sha256": "55e1a3e0085836221fb1676034c321a03cb43f7cd2a2d9e1dc513f200c40c555"
            },
            "outputSha256": "55e1a3e0085836221fb1676034c321a03cb43f7cd2a2d9e1dc513f200c40c555",
            "outputValue": 5.017837047576904,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0003/weight.npy",
              "sha256": "8c1ccb9a371ba005c5c23915a34d42c2a3e6b16cad85f33d544e67e547d9fb76"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.4285712242126465,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/output.npy",
              "sha256": "8470e9f887ee1e93fa38753228ea82e1bf266d1497767e7ace52549d53310f9c"
            },
            "outputSha256": "8470e9f887ee1e93fa38753228ea82e1bf266d1497767e7ace52549d53310f9c",
            "outputValue": 5.42857027053833,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0004/weight.npy",
              "sha256": "69aa3b5be30dbfd7991c939426d19626827444f69b553d2b484df8287e3bac72"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.3615249395370483,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/output.npy",
              "sha256": "1c5b6c6fe4b587a58d9fcfa54d07f4c97e71fcc419971735ac185976bd29ccd3"
            },
            "outputSha256": "1c5b6c6fe4b587a58d9fcfa54d07f4c97e71fcc419971735ac185976bd29ccd3",
            "outputValue": 1.3615249395370483,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-11082/chunk-0005/weight.npy",
              "sha256": "ed22edcfdc58e911b65773aebbab810f82346b59e7cbb0b5797fd6862aee2855"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 32.9051513671875,
        "cpuSoftcappedLogit": 23.980814568115942,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 24.037384033203125,
        "lmHeadRowSha256": "ad556d7bfcac36675bb71fa94622b58b27bd12e458335df8a373ba4824620cac",
        "logitAbsDiff": 0.05657084228006326,
        "rawLogit": 32.905147552490234,
        "referenceRank": 4,
        "softcappedLogit": 23.98081319092306,
        "tokenId": 11082
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 38.49152374267578,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/output.npy",
              "sha256": "9a3efd4ca87639b2f2a72668aa04c036e4345158a0983a6ce62f28d00497961c"
            },
            "outputSha256": "9a3efd4ca87639b2f2a72668aa04c036e4345158a0983a6ce62f28d00497961c",
            "outputValue": 38.49152374267578,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0000/weight.npy",
              "sha256": "0c18ec6cc379336c69fdd661dd6bc37dcd83fb239db1163693318871050b4af5"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 25.231826782226562,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/output.npy",
              "sha256": "cc1871dee79da53c3cf4cebcccdcc441825c8593c3cbd4cc86be13084f0a5345"
            },
            "outputSha256": "cc1871dee79da53c3cf4cebcccdcc441825c8593c3cbd4cc86be13084f0a5345",
            "outputValue": 25.231830596923828,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0001/weight.npy",
              "sha256": "3d029890a10cb1b195770557c349b5d01fa43e5c59d47fa124ab2907236ced14"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.8931989669799805,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/output.npy",
              "sha256": "93d885622a4f3ff5cde5709d2cb76f080a914c48ca4e2d8c190b2c1ee40fdd81"
            },
            "outputSha256": "93d885622a4f3ff5cde5709d2cb76f080a914c48ca4e2d8c190b2c1ee40fdd81",
            "outputValue": -0.8931972980499268,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0002/weight.npy",
              "sha256": "6265bf012f997109ef2bc3816c2bcf38438979ba1a92b33884ef0b29003d5d17"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -15.513075828552246,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/output.npy",
              "sha256": "d521e8128504096f3f6b7d9ca001a63dfa99171540f573d5d692ea8425b46ec2"
            },
            "outputSha256": "d521e8128504096f3f6b7d9ca001a63dfa99171540f573d5d692ea8425b46ec2",
            "outputValue": -15.51307487487793,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0003/weight.npy",
              "sha256": "fccf632af3eed16cea503349934e23ec8878b05821f0652b80ccac3cdd5596ef"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -14.305158615112305,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/output.npy",
              "sha256": "0b0b78647a7c289bfddf0174b1149393af4588553aade62368eff36f8e81f583"
            },
            "outputSha256": "0b0b78647a7c289bfddf0174b1149393af4588553aade62368eff36f8e81f583",
            "outputValue": -14.305157661437988,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0004/weight.npy",
              "sha256": "8fd6fb1df3559e44f633262c6b2ddf56089de2ca5482a02b476223ac68edabe1"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.9135034084320068,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/output.npy",
              "sha256": "b168d9539f4f25ac5249e3f0dab9c920419d54382f7b6fc45b011e8e87b4e22f"
            },
            "outputSha256": "b168d9539f4f25ac5249e3f0dab9c920419d54382f7b6fc45b011e8e87b4e22f",
            "outputValue": -0.91350257396698,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-496/chunk-0005/weight.npy",
              "sha256": "87773870be953c6a1c96b9de9e561180f94bfa15a6544c4b963cd3363fd691aa"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 32.09840393066406,
        "cpuSoftcappedLogit": 23.68323528270532,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 23.763450622558594,
        "lmHeadRowSha256": "c597008e60d1cd69ec0aae0d44bd620be1b90779907832c1621ea5d965c03b22",
        "logitAbsDiff": 0.08020671598847073,
        "rawLogit": 32.098426818847656,
        "referenceRank": 5,
        "softcappedLogit": 23.683243906570123,
        "tokenId": 496
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.781267166137695,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/output.npy",
              "sha256": "14ea0d8721c1ac553aca9f82c06cc35e4aaafe4ae7e009a9d9a69994b0151370"
            },
            "outputSha256": "14ea0d8721c1ac553aca9f82c06cc35e4aaafe4ae7e009a9d9a69994b0151370",
            "outputValue": 8.781267166137695,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0000/weight.npy",
              "sha256": "fc8abf0874db8ad15726d63dfa644a37a7a48c8ef083ffa57a26830fd56a881a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.804577827453613,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/output.npy",
              "sha256": "f10c2176974e6967cc72ee12d9c2a4e9ad46a920c69590a502ef17ad489ccab6"
            },
            "outputSha256": "f10c2176974e6967cc72ee12d9c2a4e9ad46a920c69590a502ef17ad489ccab6",
            "outputValue": 7.804577827453613,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0001/weight.npy",
              "sha256": "950a7c072a4e17d6e230526acae7ae4300c3fde19c44e3d6a40e79454afae733"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.2693610191345215,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/output.npy",
              "sha256": "aacb8f5e5e7256601055a342adbe9fd4722569d609a95194789f19287b4c92e4"
            },
            "outputSha256": "aacb8f5e5e7256601055a342adbe9fd4722569d609a95194789f19287b4c92e4",
            "outputValue": 0.26936131715774536,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0002/weight.npy",
              "sha256": "966182f667e6da929c385dda902a16ca72ae1e1ac5e1657d0e597097ece96c91"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.183527946472168,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/output.npy",
              "sha256": "c4773917eb81a949b721d6ca2141bb424348ac53a7ba73ea26dd594873e6ccf2"
            },
            "outputSha256": "c4773917eb81a949b721d6ca2141bb424348ac53a7ba73ea26dd594873e6ccf2",
            "outputValue": 5.183527946472168,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0003/weight.npy",
              "sha256": "294bc6a4ead2e7a87b58c7b1f145c89ec2e42b88e91ce4966e732e2d9d7213b9"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.366952896118164,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/output.npy",
              "sha256": "f1940ba625efa23437b7d1ccbcdfda8823c4df1bfb7db094aa2498a08ef88aa9"
            },
            "outputSha256": "f1940ba625efa23437b7d1ccbcdfda8823c4df1bfb7db094aa2498a08ef88aa9",
            "outputValue": 7.366951942443848,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0004/weight.npy",
              "sha256": "092061eb462c8917d454390b103f99fa2aa0e382b5f9351e131b7c0abdeb7109"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.7979943156242371,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/output.npy",
              "sha256": "6cef3d762b0b26be608131edc5799a1778c66c2a14c3b9b18735d007d229f615"
            },
            "outputSha256": "6cef3d762b0b26be608131edc5799a1778c66c2a14c3b9b18735d007d229f615",
            "outputValue": 0.7979943752288818,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3187/chunk-0005/weight.npy",
              "sha256": "78fd7c614ffeefdc5b9b466ac1cf3323de218dd7f80d6dc7fdeb041a992ee1ad"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 30.203683853149414,
        "cpuSoftcappedLogit": 22.932925323538036,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 22.958629608154297,
        "lmHeadRowSha256": "4623aa3406178cfed7556bc47582b3ce20cf29661a4494a882f4e8e05123f733",
        "logitAbsDiff": 0.025705870178111212,
        "rawLogit": 30.20368003845215,
        "referenceRank": 6,
        "softcappedLogit": 22.932923737976186,
        "tokenId": 3187
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.366278648376465,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/output.npy",
              "sha256": "faa3099390683c48f026d92de89a19fb9e74c023c06240d457620c84b9224ecd"
            },
            "outputSha256": "faa3099390683c48f026d92de89a19fb9e74c023c06240d457620c84b9224ecd",
            "outputValue": 8.366279602050781,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0000/weight.npy",
              "sha256": "163f4ee3fa3b14aba7b1a1a072d84d2b1d85618e71b7adf67c3ec315833004d5"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.767921447753906,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/output.npy",
              "sha256": "1d414d220a045809adb4c8d13153c04052a2ecb2fccd5db20a51e7972f13780b"
            },
            "outputSha256": "1d414d220a045809adb4c8d13153c04052a2ecb2fccd5db20a51e7972f13780b",
            "outputValue": 5.767920970916748,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0001/weight.npy",
              "sha256": "24b7befbb74d0da6fff1b27c31f1d8ad4de34e9b257338df7a91d6c652ca35a1"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.909374713897705,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/output.npy",
              "sha256": "39d7829953f4e51022b7d7420dc2c60f613a2b4ed0ab92d36b509c208a5f656a"
            },
            "outputSha256": "39d7829953f4e51022b7d7420dc2c60f613a2b4ed0ab92d36b509c208a5f656a",
            "outputValue": 4.909374237060547,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0002/weight.npy",
              "sha256": "28e0331e976160fbdea3b72c2a7605b1afe948b2aab8198a7256da63e3d19e5e"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 10.621394157409668,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/output.npy",
              "sha256": "e9ae9ff39776c2b7c746ad0f2d621fdd2734bb835c7227a584df43ec3cacd98c"
            },
            "outputSha256": "e9ae9ff39776c2b7c746ad0f2d621fdd2734bb835c7227a584df43ec3cacd98c",
            "outputValue": 10.621393203735352,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0003/weight.npy",
              "sha256": "62a2c4840e834141ac1051e12f8004c875aeb1bc19b172edc22d37aebbb31dd0"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.9615505933761597,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/output.npy",
              "sha256": "df33ec7c59658b859df940238d9c3c2c6e6b0a5f1049137dfeadcb54d0f0dd29"
            },
            "outputSha256": "df33ec7c59658b859df940238d9c3c2c6e6b0a5f1049137dfeadcb54d0f0dd29",
            "outputValue": -0.9615506529808044,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0004/weight.npy",
              "sha256": "1a75c7f3e54f11ce16bb3186d552d6365281331b53f17bc346b0c78a35efb29f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.5471563339233398,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/output.npy",
              "sha256": "10f0725a58c6085c9a8e2d665f784b831245c7ffb9d8273b36349c681f7692b1"
            },
            "outputSha256": "10f0725a58c6085c9a8e2d665f784b831245c7ffb9d8273b36349c681f7692b1",
            "outputValue": 1.5471564531326294,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2214/chunk-0005/weight.npy",
              "sha256": "8da2f4539f55bece82d7288d8cfa9ada829899da0de78ca396d967d3fad0c427"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 30.250574111938477,
        "cpuSoftcappedLogit": 22.95239177327365,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 22.94567108154297,
        "lmHeadRowSha256": "28ded977f346580b36569f0f7418b823978df8674d105bd33e134cc3c6014879",
        "logitAbsDiff": 0.00672148261850225,
        "rawLogit": 30.25057601928711,
        "referenceRank": 7,
        "softcappedLogit": 22.95239256416147,
        "tokenId": 2214
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 16.61925506591797,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/output.npy",
              "sha256": "2f24b4d0b3fb58d92cd08bcc463e900b476570f5aa95f0e54566000714331910"
            },
            "outputSha256": "2f24b4d0b3fb58d92cd08bcc463e900b476570f5aa95f0e54566000714331910",
            "outputValue": 16.61925506591797,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0000/weight.npy",
              "sha256": "3459a6e6027a60f17d487272d9ba6ff024b58436e68f8f824c83b2fbead0fa77"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -5.080806732177734,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/output.npy",
              "sha256": "33d066646df03839da5cfb6b02ada768e7c3cda2e000512032beff58a9d07c0e"
            },
            "outputSha256": "33d066646df03839da5cfb6b02ada768e7c3cda2e000512032beff58a9d07c0e",
            "outputValue": -5.080808162689209,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0001/weight.npy",
              "sha256": "df3a35e307d8839f5036d195f3b80fd962c71f880a7d1ec046cccb40d3ca4a3a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 23.890058517456055,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/output.npy",
              "sha256": "aa8b17d234acaba671b4c523800f62a20f58ed7d9da3cfa1969df2b915c5189f"
            },
            "outputSha256": "aa8b17d234acaba671b4c523800f62a20f58ed7d9da3cfa1969df2b915c5189f",
            "outputValue": 23.890058517456055,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0002/weight.npy",
              "sha256": "9a3b9f4a7542e4b31f5c580a992990d9d6c228c2366e353e5ce785a1f4b64d79"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -14.425957679748535,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/output.npy",
              "sha256": "5740a25a4c20f6ef1601821bf8f003963819cdcb1df364111faa4224a1c78db7"
            },
            "outputSha256": "5740a25a4c20f6ef1601821bf8f003963819cdcb1df364111faa4224a1c78db7",
            "outputValue": -14.425963401794434,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0003/weight.npy",
              "sha256": "941f28db5ece7d7f285715a9acb33a566222900049d369e9ff56b1d3c0a7d864"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.045406341552734375,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/output.npy",
              "sha256": "169fb6c9562f9bd8e29cb40fb7f03d314222ee41ea3706fa96ad78b6e7f7b1c8"
            },
            "outputSha256": "169fb6c9562f9bd8e29cb40fb7f03d314222ee41ea3706fa96ad78b6e7f7b1c8",
            "outputValue": -0.04540574550628662,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0004/weight.npy",
              "sha256": "8f7c2cfad994dc82a2c7af661270a4f7cde51194e05e111a11d9b165a042fcd1"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 8.01095199584961,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/output.npy",
              "sha256": "fa49ba20db88c15bb788351d9e7260e73d21129567eddb41e93e5b1a6d229bbe"
            },
            "outputSha256": "fa49ba20db88c15bb788351d9e7260e73d21129567eddb41e93e5b1a6d229bbe",
            "outputValue": 8.010951042175293,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236858/chunk-0005/weight.npy",
              "sha256": "a7afe91bd2da43e4bbdd389485e627b67a4d293f454a59bbd4c7845367418144"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 28.968109130859375,
        "cpuSoftcappedLogit": 22.402979207684645,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 22.403263092041016,
        "lmHeadRowSha256": "8a881316ace0dffd627cc559c458d5e779c93b3894ae929e1455f25e897685b2",
        "logitAbsDiff": 0.00029400873455998067,
        "rawLogit": 28.96808624267578,
        "referenceRank": 8,
        "softcappedLogit": 22.402969083306456,
        "tokenId": 236858
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.0760111808776855,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/output.npy",
              "sha256": "93549f0f119f890cc05fef5cd693defb30e93e54350b32887c1b1f949e15e781"
            },
            "outputSha256": "93549f0f119f890cc05fef5cd693defb30e93e54350b32887c1b1f949e15e781",
            "outputValue": 2.0760111808776855,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0000/weight.npy",
              "sha256": "fe4d8f0934e3f55f7622ad335b02be731ce28889e809054c822a5364973fa096"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.21842622756958,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/output.npy",
              "sha256": "a0647d77d18354ee4d823306f23da418ac03b4f960275381a2b2bcd56e3bf896"
            },
            "outputSha256": "a0647d77d18354ee4d823306f23da418ac03b4f960275381a2b2bcd56e3bf896",
            "outputValue": 4.218426704406738,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0001/weight.npy",
              "sha256": "4d51cb384644d9589dc3cbad22346164a79223ce1d03cefac27201317f07d809"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.184101104736328,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/output.npy",
              "sha256": "f358f64890c8f3d01690b2e057ea4e618877c4fecde6c2739f04b5d49d4c95b7"
            },
            "outputSha256": "f358f64890c8f3d01690b2e057ea4e618877c4fecde6c2739f04b5d49d4c95b7",
            "outputValue": 8.184103012084961,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0002/weight.npy",
              "sha256": "3e4c51dfab7529e9b3fe995fe28a7e0936efc6606de7a76d25ec99bd0da0d618"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.3057734966278076,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/output.npy",
              "sha256": "d36edb4eae80a1a6b23c12fe332242c1d8dd1c9e7b1398502fb932b405a97d7a"
            },
            "outputSha256": "d36edb4eae80a1a6b23c12fe332242c1d8dd1c9e7b1398502fb932b405a97d7a",
            "outputValue": 3.3057732582092285,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0003/weight.npy",
              "sha256": "a8c0d171c5b25d60821266098fcbbbf653df116e20faa3e020e660b2c7a12513"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.735149383544922,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/output.npy",
              "sha256": "3826c218f6a08d617601ea3b614981110d1a1a16a6f32b50ccd6b570a02a9919"
            },
            "outputSha256": "3826c218f6a08d617601ea3b614981110d1a1a16a6f32b50ccd6b570a02a9919",
            "outputValue": 7.735149383544922,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0004/weight.npy",
              "sha256": "ceb6eb0f429ac758780f9f1ca9f84d849ddc24b208ec786445168f372fc8eca5"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.4889568090438843,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/output.npy",
              "sha256": "fd584feebcd11d6bb37f8659962fad7fc1b8400a187445fb493decb925e1f18f"
            },
            "outputSha256": "fd584feebcd11d6bb37f8659962fad7fc1b8400a187445fb493decb925e1f18f",
            "outputValue": 1.4889566898345947,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14102/chunk-0005/weight.npy",
              "sha256": "18ac0bc23e0e83e7643e7166534baf625e98c2e6a5b66a36ea3479b4ead795ec"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 27.0084171295166,
        "cpuSoftcappedLogit": 21.49303372884678,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 21.594894409179688,
        "lmHeadRowSha256": "915ebb3abb4526046536e1b40bc528fc01574869c0a30096f6a3d043ab9a991f",
        "logitAbsDiff": 0.10185975198504238,
        "rawLogit": 27.008419036865234,
        "referenceRank": 9,
        "softcappedLogit": 21.493034657194645,
        "tokenId": 14102
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.2688581943511963,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/output.npy",
              "sha256": "dcb239c0cd38ae6a4b62c3246f7839781edeebf28da1324a65e6138e66c19805"
            },
            "outputSha256": "dcb239c0cd38ae6a4b62c3246f7839781edeebf28da1324a65e6138e66c19805",
            "outputValue": 3.2688589096069336,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0000/weight.npy",
              "sha256": "6c4ac495a6a32edd1228ce5fb12029784b09ff6784ed4241e982a8f74ad27beb"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.2070980072021484,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/output.npy",
              "sha256": "590b86054429dce319ba3142f7a066d42a0b77bcfcb6c14c0dbecc4605b3718f"
            },
            "outputSha256": "590b86054429dce319ba3142f7a066d42a0b77bcfcb6c14c0dbecc4605b3718f",
            "outputValue": 1.2070977687835693,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0001/weight.npy",
              "sha256": "4708b9405508a03ccaccd9728ece1b4005b6a3d93a0e003c51adb7063551683b"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.875068664550781,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/output.npy",
              "sha256": "48643f526d985fb7490e841f55e4883bf279975de7ec185c775ffd10faef3999"
            },
            "outputSha256": "48643f526d985fb7490e841f55e4883bf279975de7ec185c775ffd10faef3999",
            "outputValue": 7.8750691413879395,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0002/weight.npy",
              "sha256": "9ecc97c251cb8cbaab1293daba163c2857cb7ceb217c93254b209e7dbe9826f2"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.354341506958008,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/output.npy",
              "sha256": "2ed19d33b5001d96f02dbfc9e58e250b8cb100a3b5b47200befff19a5e4b171d"
            },
            "outputSha256": "2ed19d33b5001d96f02dbfc9e58e250b8cb100a3b5b47200befff19a5e4b171d",
            "outputValue": 6.354341983795166,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0003/weight.npy",
              "sha256": "f7d119ffba35385306e886b0af6d041d86ecb5d809491316c3d8e29a9b1f7973"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.596238613128662,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/output.npy",
              "sha256": "b156d43d9dc12dca0e6f77c1904a8f6ed61166c9fce93124dbd12d083effa2ca"
            },
            "outputSha256": "b156d43d9dc12dca0e6f77c1904a8f6ed61166c9fce93124dbd12d083effa2ca",
            "outputValue": 6.596238136291504,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0004/weight.npy",
              "sha256": "58befc4a9acce2860254ce49b42d1261cb634d0039bc0056864bce455b7dc00a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.7849223613739014,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/output.npy",
              "sha256": "8c4e8237cecc277b4c40ecfcf3d1d21c4f6baa364e9d7eddac785da9d0fe03b5"
            },
            "outputSha256": "8c4e8237cecc277b4c40ecfcf3d1d21c4f6baa364e9d7eddac785da9d0fe03b5",
            "outputValue": 0.7849223613739014,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-6816/chunk-0005/weight.npy",
              "sha256": "81bd6a618c7eb0132a7022f9f29f0b2e5f73e3631a666e49dbce8579111536e7"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 26.086528778076172,
        "cpuSoftcappedLogit": 21.034377390290544,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 21.101390838623047,
        "lmHeadRowSha256": "781d05d90c05db41305dc8b201a142cee76e4474019979c3b0c5eb4d19fed89d",
        "logitAbsDiff": 0.06701441801792996,
        "rawLogit": 26.08652687072754,
        "referenceRank": 10,
        "softcappedLogit": 21.034376420605117,
        "tokenId": 6816
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.845949172973633,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/output.npy",
              "sha256": "561749d6bd60443034e188212a4cf7efa5f688f3f40a7c00e740c254ae0be82d"
            },
            "outputSha256": "561749d6bd60443034e188212a4cf7efa5f688f3f40a7c00e740c254ae0be82d",
            "outputValue": 8.84595012664795,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0000/weight.npy",
              "sha256": "a099e08e1cdc887693ec90cc065fb2cf3d5c5cd1435687f780b1b0b8896fed08"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.8167798519134521,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/output.npy",
              "sha256": "0a65385072f2df9e59c24b2d94953fddb6890dc881712eb7bc5cc383c288a7d2"
            },
            "outputSha256": "0a65385072f2df9e59c24b2d94953fddb6890dc881712eb7bc5cc383c288a7d2",
            "outputValue": 0.8167803287506104,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0001/weight.npy",
              "sha256": "0d6c211d8b153000ec886acc551ab14af8c9106a455cb899e96b0027045fece5"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.898411750793457,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/output.npy",
              "sha256": "05c55f4c719554014c284430dc459ccbfb4a5ad17d494ebad0d4a2bb24668106"
            },
            "outputSha256": "05c55f4c719554014c284430dc459ccbfb4a5ad17d494ebad0d4a2bb24668106",
            "outputValue": 6.898410797119141,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0002/weight.npy",
              "sha256": "28b1d1a9bf3a6d4c7bf7f6795dad3a0b229534947d0bfe312f75f23b8a6f9935"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.886726379394531,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/output.npy",
              "sha256": "666c18678b3cb7e6d3b301f81ae4cff51a9b6e401d990fb60ff742e2558b8b06"
            },
            "outputSha256": "666c18678b3cb7e6d3b301f81ae4cff51a9b6e401d990fb60ff742e2558b8b06",
            "outputValue": 8.886727333068848,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0003/weight.npy",
              "sha256": "15964f83d5140a2de32ef3157cfba45b4d83576dca45c66c95ffbabfe1d0f811"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.4720015525817871,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/output.npy",
              "sha256": "018d8b521876aa725090278e0438dfa44de65da51c7018b50e5c03f8a038b883"
            },
            "outputSha256": "018d8b521876aa725090278e0438dfa44de65da51c7018b50e5c03f8a038b883",
            "outputValue": -0.4720005393028259,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0004/weight.npy",
              "sha256": "7e2b186a745bff0785e45f8c536059ff76b4b73278ac2f6ad9ca04678a5dff6f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.8551340103149414,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/output.npy",
              "sha256": "d854f1ee616cea212f781d895c85c5792ff0e2a336bbcbec27fec3b3a5d3296d"
            },
            "outputSha256": "d854f1ee616cea212f781d895c85c5792ff0e2a336bbcbec27fec3b3a5d3296d",
            "outputValue": 0.8551339507102966,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1390/chunk-0005/weight.npy",
              "sha256": "87fc7b4005cde8aa320562886afd3d8bb3b6069ba0ca7218e779473689fdafac"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.83100128173828,
        "cpuSoftcappedLogit": 20.903691334166595,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.926305770874023,
        "lmHeadRowSha256": "3a8aecb26db8c89ef37f4fcf1d1550d05f8890999336bcb9614c411c7d69392a",
        "logitAbsDiff": 0.022614436707428354,
        "rawLogit": 25.83100128173828,
        "referenceRank": 11,
        "softcappedLogit": 20.903691334166595,
        "tokenId": 1390
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.069339275360107,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/output.npy",
              "sha256": "900816d28ae608eb7737bd6298e6ebc31a6f8ac452bfe54abf14c9e6d2ef4a41"
            },
            "outputSha256": "900816d28ae608eb7737bd6298e6ebc31a6f8ac452bfe54abf14c9e6d2ef4a41",
            "outputValue": 7.069339752197266,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0000/weight.npy",
              "sha256": "1f393fbd04c18df63dbe3ad0072ac0fea332f5579cc1a173cb665e21b276ff85"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 15.416681289672852,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/output.npy",
              "sha256": "e81c90d34f2ae3c2a1aeb1449643dec6f1c91f1ab1d9b796b366747ba1b9a949"
            },
            "outputSha256": "e81c90d34f2ae3c2a1aeb1449643dec6f1c91f1ab1d9b796b366747ba1b9a949",
            "outputValue": 15.416681289672852,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0001/weight.npy",
              "sha256": "4a0427df0fcfc6f66059550c4e6410c34bfb944d3ee43b4af1fb7122753437bf"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.4286024570465088,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/output.npy",
              "sha256": "0f674e8d9cdb76a3a7deb81334ccbbb3176884a1293924df1d713ad76e51720b"
            },
            "outputSha256": "0f674e8d9cdb76a3a7deb81334ccbbb3176884a1293924df1d713ad76e51720b",
            "outputValue": 0.4286026656627655,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0002/weight.npy",
              "sha256": "6a90700130e5e07f59b18e4300796796602b746d8d188ff50d076b9e74e16b7d"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.494356155395508,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/output.npy",
              "sha256": "cfdd8958c2b654ff607be0ed79697c79f7607b2ce0420e4378f7f130f5acb083"
            },
            "outputSha256": "cfdd8958c2b654ff607be0ed79697c79f7607b2ce0420e4378f7f130f5acb083",
            "outputValue": 3.494356393814087,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0003/weight.npy",
              "sha256": "2d7d9ef351b86742f69d9f366f25362a387073f29ff01ae8b2e3b89807816954"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -2.012533664703369,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/output.npy",
              "sha256": "a0cd2aa294afbf1b80f743f051f943bbbb3ca1eeb1e33254005c04627ee7f726"
            },
            "outputSha256": "a0cd2aa294afbf1b80f743f051f943bbbb3ca1eeb1e33254005c04627ee7f726",
            "outputValue": -2.012533664703369,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0004/weight.npy",
              "sha256": "22b4a508a390e70ab316f0c8a92451e2db2bf17dfe296b7213311d542607afe9"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.3180326223373413,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/output.npy",
              "sha256": "2ad0f371d0c02e4d3eccf4990ef7098332dd7e90af832e61a413b89324b82f34"
            },
            "outputSha256": "2ad0f371d0c02e4d3eccf4990ef7098332dd7e90af832e61a413b89324b82f34",
            "outputValue": 1.3180327415466309,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-506/chunk-0005/weight.npy",
              "sha256": "9240047f2f7c31cd74a03afa4dcb1a072f7175d522a7daf99950596d52c8322c"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.714475631713867,
        "cpuSoftcappedLogit": 20.843578348935317,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.82151985168457,
        "lmHeadRowSha256": "1f7bf1bf2a9614e5ccd8e0742de914e0e4f8f952faad6db04d7951b4c7391899",
        "logitAbsDiff": 0.02206047048852966,
        "rawLogit": 25.714479446411133,
        "referenceRank": 12,
        "softcappedLogit": 20.8435803221731,
        "tokenId": 506
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.891907215118408,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/output.npy",
              "sha256": "83f66cd7bcac9bbadca0a616f006bea641ff3daf8b114e3a7e83c83e8b85ed5d"
            },
            "outputSha256": "83f66cd7bcac9bbadca0a616f006bea641ff3daf8b114e3a7e83c83e8b85ed5d",
            "outputValue": 3.8919079303741455,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0000/weight.npy",
              "sha256": "6d0c7f907a89c5e89dd6c9a86d7671ea9bdbf64b32b14a8038a83f8c87107839"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 10.423118591308594,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/output.npy",
              "sha256": "7e538839ad300a861c698ca1886fd9cf4865c8a7789fa138af830641affabfd0"
            },
            "outputSha256": "7e538839ad300a861c698ca1886fd9cf4865c8a7789fa138af830641affabfd0",
            "outputValue": 10.423118591308594,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0001/weight.npy",
              "sha256": "82705e3dca2969fea3770cbe0ecce44805f50f70d16171e6e2966e4321356e6e"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.7470005750656128,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/output.npy",
              "sha256": "297b54c7d83741d070566bed214f6865f44a27b975a7b6e8c1313fe04af69fbb"
            },
            "outputSha256": "297b54c7d83741d070566bed214f6865f44a27b975a7b6e8c1313fe04af69fbb",
            "outputValue": 1.7469998598098755,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0002/weight.npy",
              "sha256": "3adedd22862a1dc135ead3c1b51d7a583654605a7472b8110ae626152fa46ee7"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.89449405670166,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/output.npy",
              "sha256": "01eca993cadceb9f6487d67771017a2b7f343e369697d62a2fca72acb2e0e544"
            },
            "outputSha256": "01eca993cadceb9f6487d67771017a2b7f343e369697d62a2fca72acb2e0e544",
            "outputValue": 6.894493579864502,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0003/weight.npy",
              "sha256": "e31b84a6b67964fd6d113eb00a1b02f8dcfedd82c5ebf72a9c4c3b0c23b7747a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.5330188274383545,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/output.npy",
              "sha256": "78a1eb7a2e602503d7e50bb4356a41a6d7a26c7bb46fcc032b5a8196ef0de7f0"
            },
            "outputSha256": "78a1eb7a2e602503d7e50bb4356a41a6d7a26c7bb46fcc032b5a8196ef0de7f0",
            "outputValue": 1.5330193042755127,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0004/weight.npy",
              "sha256": "4badc59c99ce7dbda985cfdff6b90c8b6347834f20345741e40476a07c0d6c0f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.0698621273040771,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/output.npy",
              "sha256": "f31ead724df3339112377d89e065615793bbc6ecd0b17a7bb583f428b2d44978"
            },
            "outputSha256": "f31ead724df3339112377d89e065615793bbc6ecd0b17a7bb583f428b2d44978",
            "outputValue": 1.0698621273040771,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9595/chunk-0005/weight.npy",
              "sha256": "d42b4a65c30cfda3b9d79807b1525967d929eb405a84c1aa14265f35aba336f8"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.559402465820312,
        "cpuSoftcappedLogit": 20.76307486119659,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.780956268310547,
        "lmHeadRowSha256": "1778f7b7306eba4b567a52e6cd368b88366bb6df85a4412f7f1bac628cce91e4",
        "logitAbsDiff": 0.017881407113957692,
        "rawLogit": 25.559402465820312,
        "referenceRank": 13,
        "softcappedLogit": 20.76307486119659,
        "tokenId": 9595
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.950963020324707,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/output.npy",
              "sha256": "92272cd938858767e297460e15c5a762a70ed9e7f0fda35713b6fd888f009746"
            },
            "outputSha256": "92272cd938858767e297460e15c5a762a70ed9e7f0fda35713b6fd888f009746",
            "outputValue": 8.950963020324707,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0000/weight.npy",
              "sha256": "7f06bc316af332bbb91f2ff880aa13c7176a31466f671dc7ccb3d9b71f426de6"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.685248851776123,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/output.npy",
              "sha256": "d7fb5e0322acd9e7556f733771365c71a794c390a429faeda8a6b753d8935f06"
            },
            "outputSha256": "d7fb5e0322acd9e7556f733771365c71a794c390a429faeda8a6b753d8935f06",
            "outputValue": 7.685247898101807,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0001/weight.npy",
              "sha256": "cfbb71df1b93907c6329c4f7b9592905e7e94389f2cc49b63ffdd05113a3a3b0"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -2.1425657272338867,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/output.npy",
              "sha256": "d50f96dcf5b76ece0c0c952302047006db62740676e8e309dfbbec1401ffa660"
            },
            "outputSha256": "d50f96dcf5b76ece0c0c952302047006db62740676e8e309dfbbec1401ffa660",
            "outputValue": -2.1425657272338867,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0002/weight.npy",
              "sha256": "2e0f5b35d2ee178fa45a6dc86960163f587cf8b596b33fa6267647d107a030f4"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.258883476257324,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/output.npy",
              "sha256": "af6eb0a67c1ff9d28117113300203fc9313d6a4c5e1409ea1cebf1c4d3ab11a7"
            },
            "outputSha256": "af6eb0a67c1ff9d28117113300203fc9313d6a4c5e1409ea1cebf1c4d3ab11a7",
            "outputValue": 4.258882999420166,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0003/weight.npy",
              "sha256": "3c00b78589b63c46fa89d365e1ae815e8d3b03a1407bd2e70039808fa642509d"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.560908317565918,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/output.npy",
              "sha256": "f0b61a2387d250808f6f0a727acdd056db773c9179a90e47b82fff4f67cc0aa1"
            },
            "outputSha256": "f0b61a2387d250808f6f0a727acdd056db773c9179a90e47b82fff4f67cc0aa1",
            "outputValue": 5.560908794403076,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0004/weight.npy",
              "sha256": "f623b256f4ebd4820e26644c197157b2e6d698c501ac89de0f683032480e1c1c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.178051233291626,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/output.npy",
              "sha256": "0714d332dcdaaf08371cf62d5e6ce87de648f941631c48bb3d7a97099a7f200b"
            },
            "outputSha256": "0714d332dcdaaf08371cf62d5e6ce87de648f941631c48bb3d7a97099a7f200b",
            "outputValue": 1.1780513525009155,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1144/chunk-0005/weight.npy",
              "sha256": "01bffe5822e9663382d0fa04c41f5cc543f530d8a3b14e5c5972f3c2832f12e0"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.49148941040039,
        "cpuSoftcappedLogit": 20.727637095639412,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.756546020507812,
        "lmHeadRowSha256": "d0e7518f58c7246a455d1dd2997463975c0dc997615fcac555cfd24967c60d57",
        "logitAbsDiff": 0.02890892486840002,
        "rawLogit": 25.49148941040039,
        "referenceRank": 14,
        "softcappedLogit": 20.727637095639412,
        "tokenId": 1144
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.0312323570251465,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/output.npy",
              "sha256": "dfc7d734f2baa44bf47e44320e345cb6968cbe397447fb89ccaeafe718a662b5"
            },
            "outputSha256": "dfc7d734f2baa44bf47e44320e345cb6968cbe397447fb89ccaeafe718a662b5",
            "outputValue": 3.0312323570251465,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0000/weight.npy",
              "sha256": "a5cd02d75e85dfe1ffbf07ee2c9f29b8f0e35f2de81a55f199e1ef6c099e2f52"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.80064058303833,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/output.npy",
              "sha256": "0c5b9d7498952afd11f28b11b4a90e87c195bca492e600917a501b850aa3efbb"
            },
            "outputSha256": "0c5b9d7498952afd11f28b11b4a90e87c195bca492e600917a501b850aa3efbb",
            "outputValue": 2.800640106201172,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0001/weight.npy",
              "sha256": "10c974e5c934ad1d8a9655c70f95e0ab3bb5f43370033c9b0b9aaf25babd2afe"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.100133895874023,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/output.npy",
              "sha256": "502fdc2ea22ab9c9901604a7147f6be363d702b0a21c3f88383425913468fb7c"
            },
            "outputSha256": "502fdc2ea22ab9c9901604a7147f6be363d702b0a21c3f88383425913468fb7c",
            "outputValue": 7.100133895874023,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0002/weight.npy",
              "sha256": "d1282c5187f1a9da48a5722501459c2bd95f276e453f6819cc94a82fdf61892a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.375976085662842,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/output.npy",
              "sha256": "786343175387f3a9ddd686f6b8aa33711bb1088f6dacc6e4220e893ad20aee43"
            },
            "outputSha256": "786343175387f3a9ddd686f6b8aa33711bb1088f6dacc6e4220e893ad20aee43",
            "outputValue": 5.375975608825684,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0003/weight.npy",
              "sha256": "2a950738ba8f3e5a09cefb5f5112beee8fd9ad7e8b6de595ef0937a159dc4225"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.1819000244140625,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/output.npy",
              "sha256": "c2dcf55a7cc05978f2d9353fd90c5e36d73871a8ef5f44d5c800e0e3485bce0d"
            },
            "outputSha256": "c2dcf55a7cc05978f2d9353fd90c5e36d73871a8ef5f44d5c800e0e3485bce0d",
            "outputValue": 7.1819000244140625,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0004/weight.npy",
              "sha256": "14747d02ceb2d91a41bab182f1808635d26098674a0640f96cf4e5ffe7399df9"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.12798619270324707,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/output.npy",
              "sha256": "135253c6245c594185785e6ce4ab0c750ad19f66a7c8a030d1207a790958bec7"
            },
            "outputSha256": "135253c6245c594185785e6ce4ab0c750ad19f66a7c8a030d1207a790958bec7",
            "outputValue": -0.1279861330986023,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-9199/chunk-0005/weight.npy",
              "sha256": "2d8a94a60784443fbf23bc192a18c1b1e8edc8a6dd91e179881c48a7a015be77"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.361894607543945,
        "cpuSoftcappedLogit": 20.659704911190442,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.674800872802734,
        "lmHeadRowSha256": "003d20406f1b57f873f4c82da989c53abd3d970d8f2e625d1412c9846305a688",
        "logitAbsDiff": 0.015094958820416338,
        "rawLogit": 25.361896514892578,
        "referenceRank": 15,
        "softcappedLogit": 20.659705913982318,
        "tokenId": 9199
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 17.722795486450195,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/output.npy",
              "sha256": "f1d1688a7195f96f9f2aa69c64375b62cca3062e5f706a4ec38d9647360f0498"
            },
            "outputSha256": "f1d1688a7195f96f9f2aa69c64375b62cca3062e5f706a4ec38d9647360f0498",
            "outputValue": 17.722797393798828,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0000/weight.npy",
              "sha256": "314152e545942788679b26432b9664f2673f4981dd3971f4d45393dda9571e1f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 18.6218318939209,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/output.npy",
              "sha256": "2e0aaa3968e53396e69a53152d7832e4aa8e8b7293ded60761609b203ab5791e"
            },
            "outputSha256": "2e0aaa3968e53396e69a53152d7832e4aa8e8b7293ded60761609b203ab5791e",
            "outputValue": 18.6218318939209,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0001/weight.npy",
              "sha256": "279915fdcd553f0187527abe0d4c17896d4ecc0ff133af6725e29e2665220550"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.623687744140625,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/output.npy",
              "sha256": "b43f011b5e89273e9fdffefec37c6679c09bc7ee7ee1a1f8b6d63616e5247806"
            },
            "outputSha256": "b43f011b5e89273e9fdffefec37c6679c09bc7ee7ee1a1f8b6d63616e5247806",
            "outputValue": 2.6236863136291504,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0002/weight.npy",
              "sha256": "f28aa8072a4c49f73d0326105c7d53c752b9bceafddd274c590608b4b1cb2f35"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -10.411978721618652,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/output.npy",
              "sha256": "44b1d7f33f13ee11cff8552d5d61a1286aa58c25a546066811ca85d67356c8e1"
            },
            "outputSha256": "44b1d7f33f13ee11cff8552d5d61a1286aa58c25a546066811ca85d67356c8e1",
            "outputValue": -10.411979675292969,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0003/weight.npy",
              "sha256": "18ab12364bbcd3f593aafef225f1525760e20170c57b26e8282dc2ba77ac7d59"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -3.291019916534424,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/output.npy",
              "sha256": "e1ba5fae522d20454fdce1f1fb4dcf6f5983d7412b7758f1d6ccf78c3c7068ba"
            },
            "outputSha256": "e1ba5fae522d20454fdce1f1fb4dcf6f5983d7412b7758f1d6ccf78c3c7068ba",
            "outputValue": -3.2910208702087402,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0004/weight.npy",
              "sha256": "187762609cf9e249b318a9c0d18baa1682196e5023f6349e742ab6b144f53371"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.06405484676361084,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/output.npy",
              "sha256": "947dbe574374c8fdc07efd76686afe7fa54f04a1980a5d2b25eb27109b566246"
            },
            "outputSha256": "947dbe574374c8fdc07efd76686afe7fa54f04a1980a5d2b25eb27109b566246",
            "outputValue": -0.06405481696128845,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-614/chunk-0005/weight.npy",
              "sha256": "eaf2d80db7d2f2684fde865316ca43d9a74f3f2730db7dfdb8f9494c2b43173f"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.20126724243164,
        "cpuSoftcappedLogit": 20.574943062569503,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.584197998046875,
        "lmHeadRowSha256": "f6bfb92d6788249b45584dbdd3785f7cd7426e45f9a1caf95251dc37626d8844",
        "logitAbsDiff": 0.00925897627429606,
        "rawLogit": 25.20125961303711,
        "referenceRank": 16,
        "softcappedLogit": 20.57493902177258,
        "tokenId": 614
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.653789520263672,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/output.npy",
              "sha256": "aa678e31a6b262ca4d19382b7bea92509ea7c17ac505f5b61edc92d5d4fd2abb"
            },
            "outputSha256": "aa678e31a6b262ca4d19382b7bea92509ea7c17ac505f5b61edc92d5d4fd2abb",
            "outputValue": 6.653789520263672,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0000/weight.npy",
              "sha256": "9471835ee1d264bcdf1420415bf955a0cdedce24830237be23d081564b0c657a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.295463562011719,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/output.npy",
              "sha256": "c45ac9f5ad251acef361e869867b5b401d8bb238fa6df153e5fd869501653bb1"
            },
            "outputSha256": "c45ac9f5ad251acef361e869867b5b401d8bb238fa6df153e5fd869501653bb1",
            "outputValue": 4.295463562011719,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0001/weight.npy",
              "sha256": "1199d74a9876bb4e1a3a415f8ccf3806599e65198c388e4a8343a79b6181e6e9"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.271461486816406,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/output.npy",
              "sha256": "fc708ce06e619dd11f56cd1277a983d11d5dfce54ca3692773d706d6097286aa"
            },
            "outputSha256": "fc708ce06e619dd11f56cd1277a983d11d5dfce54ca3692773d706d6097286aa",
            "outputValue": 5.271461486816406,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0002/weight.npy",
              "sha256": "29ba54c26dd0a299c16a28e4458bbcf375dcb4e2bb02bb3a6bcd3f52a7d4c95b"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.9805307388305664,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/output.npy",
              "sha256": "60c7be07dde7d5befc7f872ddf50e6a48f2e69366964705e0b1507aeb98004d6"
            },
            "outputSha256": "60c7be07dde7d5befc7f872ddf50e6a48f2e69366964705e0b1507aeb98004d6",
            "outputValue": 0.9805310964584351,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0003/weight.npy",
              "sha256": "19a216b503d2788ec4c5cb2a2b6579d2dec727210e6e0aa954b1f8f0f6842287"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.706326007843018,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/output.npy",
              "sha256": "918350187d5b7530be8b342e07119e9ce14f589939d999319a9748923ab61bf4"
            },
            "outputSha256": "918350187d5b7530be8b342e07119e9ce14f589939d999319a9748923ab61bf4",
            "outputValue": 6.706326961517334,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0004/weight.npy",
              "sha256": "94e3206f18d102f9a1be02fdcfcccef9043d74abc7bbbebd80105aea48fbdf47"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.217769980430603,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/output.npy",
              "sha256": "c0c88a19476c0cd89ce8d6ea48d2e9fb12d2b407f7a68373bf2990ab4a8f32b2"
            },
            "outputSha256": "c0c88a19476c0cd89ce8d6ea48d2e9fb12d2b407f7a68373bf2990ab4a8f32b2",
            "outputValue": 1.217769980430603,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-37155/chunk-0005/weight.npy",
              "sha256": "ac029452fb7601e7ba007bb841e070283e30b615d5e29508c070ebbf4deb34a2"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 25.125343322753906,
        "cpuSoftcappedLogit": 20.534661247823273,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.509538650512695,
        "lmHeadRowSha256": "2505112d2cf6eb980974dbee2af93a20470ab52352237800283227c66efc5996",
        "logitAbsDiff": 0.025122597310577532,
        "rawLogit": 25.125343322753906,
        "referenceRank": 17,
        "softcappedLogit": 20.534661247823273,
        "tokenId": 37155
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.493572235107422,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/output.npy",
              "sha256": "8b2cf81c84dc9441fda94447f4252d94bd5fc7bb5253d834eac378d2645ed64a"
            },
            "outputSha256": "8b2cf81c84dc9441fda94447f4252d94bd5fc7bb5253d834eac378d2645ed64a",
            "outputValue": 8.493570327758789,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0000/weight.npy",
              "sha256": "c746669e6d5a5876217eb49b836ecf80298053f997d9bf2000ad190612741816"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.172407627105713,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/output.npy",
              "sha256": "794cf184ec1eee12a62c6d9e22f898c6d6f4ad309ad11b7b5de421525641ce8a"
            },
            "outputSha256": "794cf184ec1eee12a62c6d9e22f898c6d6f4ad309ad11b7b5de421525641ce8a",
            "outputValue": 6.172406196594238,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0001/weight.npy",
              "sha256": "4897bd4ced6d3cd2aa316845d43ec80d90aa2d75c6cab77d0b4d54e7d7551436"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.0257039070129395,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/output.npy",
              "sha256": "7405ca74c65e598014ceb5f51f874026845558c61f7275b0bb56ce0d13385aea"
            },
            "outputSha256": "7405ca74c65e598014ceb5f51f874026845558c61f7275b0bb56ce0d13385aea",
            "outputValue": 4.0257039070129395,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0002/weight.npy",
              "sha256": "ed9435efb057700a0184a8be6118f6cc5b9acdcd4a9fb38cca4a5bbd2f5d9eea"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.7019872665405273,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/output.npy",
              "sha256": "4511f1dec09700add3616b563ec266e990a61b33bc97fa683ac5cc24cb0f0937"
            },
            "outputSha256": "4511f1dec09700add3616b563ec266e990a61b33bc97fa683ac5cc24cb0f0937",
            "outputValue": 3.701986312866211,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0003/weight.npy",
              "sha256": "fad09f7dd28e961437d3c738f1f1081c29e1ba8f551dad4156e252a54ce0dc1c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.7852163314819336,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/output.npy",
              "sha256": "33932c3b86de23d61e8d84ffd9bac2136e917fc844ec6f5886cd8be784cb5c45"
            },
            "outputSha256": "33932c3b86de23d61e8d84ffd9bac2136e917fc844ec6f5886cd8be784cb5c45",
            "outputValue": 2.785216808319092,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0004/weight.npy",
              "sha256": "e956be646d8816256119e2cd566575d82848f9d9d9048d1652f739055de4c3cc"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.3125237822532654,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/output.npy",
              "sha256": "2f391d1e16da5cf855741d57245098896327e3b3ac0ef443c0e556269adf3256"
            },
            "outputSha256": "2f391d1e16da5cf855741d57245098896327e3b3ac0ef443c0e556269adf3256",
            "outputValue": -0.31252387166023254,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3643/chunk-0005/weight.npy",
              "sha256": "c8f26578532cf49e60cb04d0b5008bc83eb5c2b63daa636a2761486c0c312e37"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 24.866363525390625,
        "cpuSoftcappedLogit": 20.396205214526653,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.45113754272461,
        "lmHeadRowSha256": "f39ee9a470c8c9011ac674939355c2fe4f93c49ca95c721a3e775991de2dcce4",
        "logitAbsDiff": 0.05493437963556502,
        "rawLogit": 24.86635971069336,
        "referenceRank": 18,
        "softcappedLogit": 20.396203163089044,
        "tokenId": 3643
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 9.481945037841797,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/output.npy",
              "sha256": "caba6562da5206637ad74ab35f842d28b1aea6e5d3dc9562f7477488b24eb955"
            },
            "outputSha256": "caba6562da5206637ad74ab35f842d28b1aea6e5d3dc9562f7477488b24eb955",
            "outputValue": 9.48194408416748,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0000/weight.npy",
              "sha256": "63c21142f320d349f97b2c6528a40ab14141b1826dff02bab6d6453a564a2a61"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.5294218063354492,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/output.npy",
              "sha256": "8a4d403370f1a9ff3e5998bde5b6eea4d60b6f5b0576db0491892bc731823080"
            },
            "outputSha256": "8a4d403370f1a9ff3e5998bde5b6eea4d60b6f5b0576db0491892bc731823080",
            "outputValue": 1.5294218063354492,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0001/weight.npy",
              "sha256": "368577e795f21c5b3b4425442b19aed0a20a5bc8895b1d78dfc2f01c326a8141"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.507993698120117,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/output.npy",
              "sha256": "f5683af1f5f06c09c10bb04a439028f7277f8abdfbd6d4f573ab3b502dc6a64c"
            },
            "outputSha256": "f5683af1f5f06c09c10bb04a439028f7277f8abdfbd6d4f573ab3b502dc6a64c",
            "outputValue": 4.50799560546875,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0002/weight.npy",
              "sha256": "a7e2d36fc2766ff5495c89203eba94d674ea1a1e30e7fa44f37e6fab167aff19"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.4525856971740723,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/output.npy",
              "sha256": "07d2df8d18fcbfc60519873f776a2bf76336d8943827fa1b809770851cb57a29"
            },
            "outputSha256": "07d2df8d18fcbfc60519873f776a2bf76336d8943827fa1b809770851cb57a29",
            "outputValue": 2.452585220336914,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0003/weight.npy",
              "sha256": "541c58f630d17581bad5eed603992d2923eecb8c0ec0335d03ac0c9cfe0ac942"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.4363932609558105,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/output.npy",
              "sha256": "84527821a9ef0a9916b6de93f8edb70724841cb573002e4e5858fd8e50712477"
            },
            "outputSha256": "84527821a9ef0a9916b6de93f8edb70724841cb573002e4e5858fd8e50712477",
            "outputValue": 5.4363932609558105,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0004/weight.npy",
              "sha256": "9ae79f6c9cd76ade8d42d2fc3633024d6df948e0d96390ccb34362a3de7e8d83"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 1.2551772594451904,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/output.npy",
              "sha256": "50dd2c4652c8d24fd2421d034723b2ff16ecbf6586a1d993333f38171fd9d5d2"
            },
            "outputSha256": "50dd2c4652c8d24fd2421d034723b2ff16ecbf6586a1d993333f38171fd9d5d2",
            "outputValue": 1.2551770210266113,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-5633/chunk-0005/weight.npy",
              "sha256": "ee56891ee3808f6c434ed282fffec6cd8eb9c1f91942f23c5141a4e7541d34d3"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 24.663516998291016,
        "cpuSoftcappedLogit": 20.286617926423112,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.367202758789062,
        "lmHeadRowSha256": "d469ea4e7b6248a5124935fa3d29814621740c86e19891997657bb24b5e4506a",
        "logitAbsDiff": 0.08058483236595038,
        "rawLogit": 24.663516998291016,
        "referenceRank": 19,
        "softcappedLogit": 20.286617926423112,
        "tokenId": 5633
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.236745834350586,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/output.npy",
              "sha256": "c9a5431e217afe5a14af54124e4b23568233689ae897763b74ef477b11413130"
            },
            "outputSha256": "c9a5431e217afe5a14af54124e4b23568233689ae897763b74ef477b11413130",
            "outputValue": 8.236745834350586,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0000/weight.npy",
              "sha256": "d06219fb1abfd93dd51bc2d4f79dca49bd587c5c6e28173b3ed32db1546b052f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.9755792617797852,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/output.npy",
              "sha256": "d71dfb7f4ea541ad2dbd0c26c974cd7b27a1dae4ee0a953c1089c622431b7475"
            },
            "outputSha256": "d71dfb7f4ea541ad2dbd0c26c974cd7b27a1dae4ee0a953c1089c622431b7475",
            "outputValue": 0.9755793809890747,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0001/weight.npy",
              "sha256": "243c93bf0b991abc5a8df43b4276258f92f5c3d6f7bb1ebffc1159945c3b58f7"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.09524393081665,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/output.npy",
              "sha256": "8b4a897b173f5bcb3736599422a2202ee6c8cb11c19397b2b44c8b730de1d66b"
            },
            "outputSha256": "8b4a897b173f5bcb3736599422a2202ee6c8cb11c19397b2b44c8b730de1d66b",
            "outputValue": 4.095243453979492,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0002/weight.npy",
              "sha256": "fbc265081e0b764b1dc4bec87e69df2e03c1ed4ac41d56d3e16999388e6cd8f8"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.944700241088867,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/output.npy",
              "sha256": "e5649b09aa2ef2168336cfc1939ef19074fc0f16d553d6c3c133b8135326ed4a"
            },
            "outputSha256": "e5649b09aa2ef2168336cfc1939ef19074fc0f16d553d6c3c133b8135326ed4a",
            "outputValue": 4.944699764251709,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0003/weight.npy",
              "sha256": "0609ce45a0dc89e9004065fc4bd56fd76a452ef9db2d66019ae6440c710674e1"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.445934772491455,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/output.npy",
              "sha256": "722e7891f9c80a6e663f3ab33483f2d2f0a0535b338ec922d84171a68d058d91"
            },
            "outputSha256": "722e7891f9c80a6e663f3ab33483f2d2f0a0535b338ec922d84171a68d058d91",
            "outputValue": 6.445934772491455,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0004/weight.npy",
              "sha256": "5c8718b8049e8640c309564d5abc8b4c525c145c6238a3519231874e268b806d"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.08280885219573975,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/output.npy",
              "sha256": "ca279d69747856771189d740353cea35f91716a601a918398aa53d571cd150d0"
            },
            "outputSha256": "ca279d69747856771189d740353cea35f91716a601a918398aa53d571cd150d0",
            "outputValue": -0.08280879259109497,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1607/chunk-0005/weight.npy",
              "sha256": "a6407bb21b4ccfd00ae8b2154650e32e0ca2279e7827e75b9b4d6e5c94e39d6a"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 24.615394592285156,
        "cpuSoftcappedLogit": 20.260472321743585,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 20.291152954101562,
        "lmHeadRowSha256": "5f817cf5c7afaa7c69d7dbefc3cde32aa47d191cd5c15705231b332ea43fc39e",
        "logitAbsDiff": 0.03068166977185527,
        "rawLogit": 24.615392684936523,
        "referenceRank": 20,
        "softcappedLogit": 20.260471284329707,
        "tokenId": 1607
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.2644581198692322,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/output.npy",
              "sha256": "1911939a7bd97de5f1209096e8db731ed990e2eaa1398c3c60add72be4176680"
            },
            "outputSha256": "1911939a7bd97de5f1209096e8db731ed990e2eaa1398c3c60add72be4176680",
            "outputValue": 0.2644582986831665,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0000/weight.npy",
              "sha256": "f5b406170f87f596b301f90d304e95a36f27832a1e06bb74deb678cc41f42a09"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.518560409545898,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/output.npy",
              "sha256": "fa1c4c1a801880ef667c839f47a7a29547bb32b326fc3fe38ce08d18550db547"
            },
            "outputSha256": "fa1c4c1a801880ef667c839f47a7a29547bb32b326fc3fe38ce08d18550db547",
            "outputValue": 7.51855993270874,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0001/weight.npy",
              "sha256": "710cea5400d707a48333a8928b417b1b2302b994b566481c98e5d73c7c43bc37"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.183831214904785,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/output.npy",
              "sha256": "2be947ea861441fd44008badf91a8bdab7a5b0f6ddfad728b73ca5a5434d465a"
            },
            "outputSha256": "2be947ea861441fd44008badf91a8bdab7a5b0f6ddfad728b73ca5a5434d465a",
            "outputValue": 2.1838326454162598,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0002/weight.npy",
              "sha256": "045537a14fb48af3cfc8be3759af170b20fbb8bb73e90403a939d616a4e68644"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.318199157714844,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/output.npy",
              "sha256": "7e91f50c7b44c30ad7e912ef08bbe807f9810d18ccfe60f920d9fca753dbabb6"
            },
            "outputSha256": "7e91f50c7b44c30ad7e912ef08bbe807f9810d18ccfe60f920d9fca753dbabb6",
            "outputValue": 8.318197250366211,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0003/weight.npy",
              "sha256": "3334013f06e4acc966f5d775992c3546d6439b66b95e2e789096981dc2c98884"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.902186393737793,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/output.npy",
              "sha256": "a4c41ee17d125d54d03f85268ab5eabc388045a16b8d46065e3ef4a9405e62bb"
            },
            "outputSha256": "a4c41ee17d125d54d03f85268ab5eabc388045a16b8d46065e3ef4a9405e62bb",
            "outputValue": 4.902186393737793,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0004/weight.npy",
              "sha256": "174d6cfa17835d1a1668711bfd16a15074055f8f96c21399b65a1dce259b0576"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.5247231721878052,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/output.npy",
              "sha256": "4018b1fcfad811dc6a3038cb8bce24df4b2f24613f0f4bf689611d6a9c34636e"
            },
            "outputSha256": "4018b1fcfad811dc6a3038cb8bce24df4b2f24613f0f4bf689611d6a9c34636e",
            "outputValue": 0.5247229933738708,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7205/chunk-0005/weight.npy",
              "sha256": "6327976852b5b214e982a91f4f95e9a47453ced4f4a16198c8aaa7b131e38842"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 23.711956024169922,
        "cpuSoftcappedLogit": 19.759042359822015,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.78056526184082,
        "lmHeadRowSha256": "d9bb31e2ebf1f6b472070ee42555c3a92dcfc240f6d345d40348ed8f28b989c3",
        "logitAbsDiff": 0.021522902018805468,
        "rawLogit": 23.711956024169922,
        "referenceRank": 21,
        "softcappedLogit": 19.759042359822015,
        "tokenId": 7205
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 11.30183219909668,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/output.npy",
              "sha256": "a4a7d38ae616f7dbd2c89baa455dc804ca5eb38bc7c4a8ba909bab4daa7135dc"
            },
            "outputSha256": "a4a7d38ae616f7dbd2c89baa455dc804ca5eb38bc7c4a8ba909bab4daa7135dc",
            "outputValue": 11.301831245422363,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0000/weight.npy",
              "sha256": "495aa4727203760e2a984e33e34e6cb098bcb63ae27b5ce20d2a74d2c4b85e5a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 11.973523139953613,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/output.npy",
              "sha256": "a19072856a9aa501be67840bc23cd3ba6f75883c15a5ebd17eb4868fe7100547"
            },
            "outputSha256": "a19072856a9aa501be67840bc23cd3ba6f75883c15a5ebd17eb4868fe7100547",
            "outputValue": 11.97352409362793,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0001/weight.npy",
              "sha256": "e585e08c4341af65c8b7fc2121595b41fb14d9b868e495397cbc8de7447b8cca"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.327162742614746,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/output.npy",
              "sha256": "e162657dae6915296240c4b99c2d1b17d6ca4c2f5508496005f26fe634264b95"
            },
            "outputSha256": "e162657dae6915296240c4b99c2d1b17d6ca4c2f5508496005f26fe634264b95",
            "outputValue": 8.327162742614746,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0002/weight.npy",
              "sha256": "819fb3d52846a4c01f9763033d945f00f104009a76d4906f4e36ca8c47abd44c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.986018657684326,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/output.npy",
              "sha256": "d8aececcefa55d82a85579b3c135254b272e56df650ad59aee32e7b3e6e7a1e6"
            },
            "outputSha256": "d8aececcefa55d82a85579b3c135254b272e56df650ad59aee32e7b3e6e7a1e6",
            "outputValue": 6.986019134521484,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0003/weight.npy",
              "sha256": "7a40e145935a1dd8ce4e0ada201c56053339586096b89e13ef3c6d8b72014ef8"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -14.186141967773438,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/output.npy",
              "sha256": "4d3e7c22bb33891d91138f2e223d56fc25944a0da083f53b1342a2452c9d171b"
            },
            "outputSha256": "4d3e7c22bb33891d91138f2e223d56fc25944a0da083f53b1342a2452c9d171b",
            "outputValue": -14.186142921447754,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0004/weight.npy",
              "sha256": "f0bcdb6715510b9cf9c66c5a3a10f04e5a833d9991156ba57b3f01029ec07f3c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.8966424465179443,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/output.npy",
              "sha256": "a9506746a33bf3055c5ae655d130947a96e3f58f2e22ae244f61e5de90fe4942"
            },
            "outputSha256": "a9506746a33bf3055c5ae655d130947a96e3f58f2e22ae244f61e5de90fe4942",
            "outputValue": -0.8966425657272339,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236743/chunk-0005/weight.npy",
              "sha256": "6a31d497f4b3aeac32f620206eccbe5116028b2c763eae7a665744c8322a4d9c"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 23.505748748779297,
        "cpuSoftcappedLogit": 19.641758627330265,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.602706909179688,
        "lmHeadRowSha256": "1c473edc5651775658fad0710cf6cc8294c8a0885677f3d262298c0fd3dcff59",
        "logitAbsDiff": 0.039054987354106174,
        "rawLogit": 23.505754470825195,
        "referenceRank": 22,
        "softcappedLogit": 19.641761896533794,
        "tokenId": 236743
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 22.260221481323242,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/output.npy",
              "sha256": "2c9f67d16716890d229b6732e34bec9a7a4a54233b69a0aa868fa40b7340d36c"
            },
            "outputSha256": "2c9f67d16716890d229b6732e34bec9a7a4a54233b69a0aa868fa40b7340d36c",
            "outputValue": 22.26021957397461,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0000/weight.npy",
              "sha256": "2efc52b22f182c219caa12314a677928a06ad1045cf7e88ad0508c7615b158b4"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -2.980785369873047,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/output.npy",
              "sha256": "07923e2a01eee30bfc7be9c5a5bef09ac6675cb41e5e18eb538e9bdaa59d9cd1"
            },
            "outputSha256": "07923e2a01eee30bfc7be9c5a5bef09ac6675cb41e5e18eb538e9bdaa59d9cd1",
            "outputValue": -2.980787515640259,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0001/weight.npy",
              "sha256": "b117dda23b03634e485f0c6c282d8dd305649fafadb7d5712c65a5be2d5ea252"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 12.06259536743164,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/output.npy",
              "sha256": "00f56b9ead59688ec856e57f60957d48eb1264acfd2a00d59edac38e3e9ddb1a"
            },
            "outputSha256": "00f56b9ead59688ec856e57f60957d48eb1264acfd2a00d59edac38e3e9ddb1a",
            "outputValue": 12.062594413757324,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0002/weight.npy",
              "sha256": "fd51fa3f222c1cb3ce0af31c92020caca1f119e069271e43ac666be12b8aafab"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -4.063118934631348,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/output.npy",
              "sha256": "8865ec5ef0335f05d10130f212c5515808f34324878ce7a0ab8f80703db21058"
            },
            "outputSha256": "8865ec5ef0335f05d10130f212c5515808f34324878ce7a0ab8f80703db21058",
            "outputValue": -4.0631208419799805,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0003/weight.npy",
              "sha256": "d63af5a7a348092c1e186ff37f079d3da42722f2f17e0c07ed7e9d4566032d16"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -9.273674011230469,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/output.npy",
              "sha256": "f477e206f5c6308e90f346eb07200534dfa7b60e185e906154df9cdad0688ea1"
            },
            "outputSha256": "f477e206f5c6308e90f346eb07200534dfa7b60e185e906154df9cdad0688ea1",
            "outputValue": -9.273672103881836,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0004/weight.npy",
              "sha256": "b5b55d94d975fe59399755b326ccf0075639a3fc6dffe9bd82d6c9ceaf5a4267"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 5.129527568817139,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/output.npy",
              "sha256": "0e6778320d89c989e3eac1d45e37c90f43049905b99839710136917b32ed2ed0"
            },
            "outputSha256": "0e6778320d89c989e3eac1d45e37c90f43049905b99839710136917b32ed2ed0",
            "outputValue": 5.129528045654297,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236789/chunk-0005/weight.npy",
              "sha256": "4ed9f1de0a4199e3209ccbd04963467f3fc243fcccf57a02d4813c501e0724a9"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 23.1347599029541,
        "cpuSoftcappedLogit": 19.42808063282537,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.458667755126953,
        "lmHeadRowSha256": "98d8d5f8a86440666160bc06f3e9bcdecc4d9ec2db4559cdc6a24994bc7a3519",
        "logitAbsDiff": 0.03058712230158278,
        "rawLogit": 23.1347599029541,
        "referenceRank": 23,
        "softcappedLogit": 19.42808063282537,
        "tokenId": 236789
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 39.77974319458008,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/output.npy",
              "sha256": "6b4093152ca1bae5140b67ab10e3d07cff383475dccabbc4ddfadce901b56bd5"
            },
            "outputSha256": "6b4093152ca1bae5140b67ab10e3d07cff383475dccabbc4ddfadce901b56bd5",
            "outputValue": 39.77973937988281,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0000/weight.npy",
              "sha256": "1b76d2e62352cc3dbef0b2f370395e1638329fc8551b472a1a92dfcc465319d8"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 11.316289901733398,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/output.npy",
              "sha256": "602b036a979f50b6f864bb6b6916553b7f4ca8dde670dca5c9a125d61f8167da"
            },
            "outputSha256": "602b036a979f50b6f864bb6b6916553b7f4ca8dde670dca5c9a125d61f8167da",
            "outputValue": 11.316291809082031,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0001/weight.npy",
              "sha256": "c5eec4e77b6b35955bd0dc73b0f65aae9db52028d3a12003d8dcc3f0d68c046e"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.0435733795166016,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/output.npy",
              "sha256": "5ab58f02bd5062ee5a08d1b6f62448d2904b5dc8bad4d72d7aadfc5578044aae"
            },
            "outputSha256": "5ab58f02bd5062ee5a08d1b6f62448d2904b5dc8bad4d72d7aadfc5578044aae",
            "outputValue": 3.043574810028076,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0002/weight.npy",
              "sha256": "1ddd9394f868a5c6dbc95daf53831770a63f53fd17ae77a4d54b90b7436ed2c7"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.112194061279297,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/output.npy",
              "sha256": "7bc42664bc999abe15de0ec1c19c2542e882ba2590e3ae9eeab51ff52e978fa5"
            },
            "outputSha256": "7bc42664bc999abe15de0ec1c19c2542e882ba2590e3ae9eeab51ff52e978fa5",
            "outputValue": 6.112195014953613,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0003/weight.npy",
              "sha256": "2db8d2e51c77a2a377b1b9cfc8447927ab8f30899c7aa4a54afda4c1f3157b31"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -43.973854064941406,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/output.npy",
              "sha256": "9644d5fcdd758d76de600be91458e42d63724e4828596e9f6e2b94d94c8b4ea6"
            },
            "outputSha256": "9644d5fcdd758d76de600be91458e42d63724e4828596e9f6e2b94d94c8b4ea6",
            "outputValue": -43.97386169433594,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0004/weight.npy",
              "sha256": "6dab77d3dfea1c4fb0aa91bfa1532ae5dc81d8a6d1f22a28e90c61265e7f72e0"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 7.007934093475342,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/output.npy",
              "sha256": "c2c5115d59526c532de7d2f69063e5f7342cc98f1fddbb5336a3a00a71fab594"
            },
            "outputSha256": "c2c5115d59526c532de7d2f69063e5f7342cc98f1fddbb5336a3a00a71fab594",
            "outputValue": 7.007932662963867,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-236772/chunk-0005/weight.npy",
              "sha256": "30367b796fa45aec01af5bd82fc24ad7edc3cd6da43062f026b134671da40b05"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 23.285884857177734,
        "cpuSoftcappedLogit": 19.515539351015025,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.40424919128418,
        "lmHeadRowSha256": "95442643b6799ee12d22cc765335c8d5cdee1e034e463e923cf90bdd174db621",
        "logitAbsDiff": 0.11128245826583338,
        "rawLogit": 23.285871505737305,
        "referenceRank": 24,
        "softcappedLogit": 19.515531649550013,
        "tokenId": 236772
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.648543357849121,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/output.npy",
              "sha256": "4cf83d774ffaf6ed50b888199847e4a460ad9bb7cb0a6fb943e67edb047d4f01"
            },
            "outputSha256": "4cf83d774ffaf6ed50b888199847e4a460ad9bb7cb0a6fb943e67edb047d4f01",
            "outputValue": 6.6485443115234375,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0000/weight.npy",
              "sha256": "e9060dfec1c1f45c99f835d34e222a337b5c71feea9cac199c1f46b508b55565"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.831256628036499,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/output.npy",
              "sha256": "b18daeee907655bfe43558311b1665c3a116d6841686f4076a731fceb5729701"
            },
            "outputSha256": "b18daeee907655bfe43558311b1665c3a116d6841686f4076a731fceb5729701",
            "outputValue": 2.831256866455078,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0001/weight.npy",
              "sha256": "7d285c7a75c743a2944074268d51216730e7d625859af9b5b7fe24aa290344fb"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.221343994140625,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/output.npy",
              "sha256": "39369490ec3a38083f50146b744f4c3cfd2369e1066835b41c97b9a0e2dc3302"
            },
            "outputSha256": "39369490ec3a38083f50146b744f4c3cfd2369e1066835b41c97b9a0e2dc3302",
            "outputValue": -0.22134345769882202,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0002/weight.npy",
              "sha256": "9a7924a51ac8b643452419fee33f6822af8940fb6a0cffc430991b2e483100dd"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.373694896697998,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/output.npy",
              "sha256": "8e931e2c3260eb34669d388d505a1d794a5dea8a7de074c40f5389cfd54aea5e"
            },
            "outputSha256": "8e931e2c3260eb34669d388d505a1d794a5dea8a7de074c40f5389cfd54aea5e",
            "outputValue": 7.37369441986084,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0003/weight.npy",
              "sha256": "98bf43c6e89cd05cd199f99923ac70f059d0467f35726b826f7d7e388c405aff"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.50376033782959,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/output.npy",
              "sha256": "48f5a5981572fc51032a998dbf5fd6212311a9e21a3419a47d3d3b36341430fb"
            },
            "outputSha256": "48f5a5981572fc51032a998dbf5fd6212311a9e21a3419a47d3d3b36341430fb",
            "outputValue": 6.503759860992432,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0004/weight.npy",
              "sha256": "85d44cb11ecbc93ce8ce3d9503000636e2e2e026f4e92fb713a98b4290a34812"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.12820160388946533,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/output.npy",
              "sha256": "820d0f8286b226b19d71456e3664ddd3dc57816af4da3372be0830120f7db8be"
            },
            "outputSha256": "820d0f8286b226b19d71456e3664ddd3dc57816af4da3372be0830120f7db8be",
            "outputValue": -0.12820172309875488,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-1401/chunk-0005/weight.npy",
              "sha256": "76e5406409a81a2e69088604cc580c7a8364066d4897d20a0dbe15954e9a93ec"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 23.007709503173828,
        "cpuSoftcappedLogit": 19.354111377498732,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.376956939697266,
        "lmHeadRowSha256": "e2874fa188a1a771be5c2c8589c266b8b3193eda838504bfc95c8e46fdbc0d2a",
        "logitAbsDiff": 0.022845562198533287,
        "rawLogit": 23.007709503173828,
        "referenceRank": 25,
        "softcappedLogit": 19.354111377498732,
        "tokenId": 1401
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -0.7010078430175781,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/output.npy",
              "sha256": "18dc4a6b5365a3795fa2a567d8f4a92b012d4ce806779eacaba9f43f73a0a514"
            },
            "outputSha256": "18dc4a6b5365a3795fa2a567d8f4a92b012d4ce806779eacaba9f43f73a0a514",
            "outputValue": -0.7010079622268677,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0000/weight.npy",
              "sha256": "6b48d43c3ed0fe79458ac73d6a41ab3484d7941a36705a32c88954be134c66e4"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.920714378356934,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/output.npy",
              "sha256": "9735a661fc885fa0c89ef07cd24aebcf2fea105ff3137c096843140f6e86dc5d"
            },
            "outputSha256": "9735a661fc885fa0c89ef07cd24aebcf2fea105ff3137c096843140f6e86dc5d",
            "outputValue": 4.920713901519775,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0001/weight.npy",
              "sha256": "11bf99058f6b18eb22accc85a766fcf86a8c85c515f64672262fa8c37048dd0c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.8321266770362854,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/output.npy",
              "sha256": "cdce7c4707b6371b4f86c288ddc8f59c622ccfc8df7c82cacb0d8cf6fac1c9d5"
            },
            "outputSha256": "cdce7c4707b6371b4f86c288ddc8f59c622ccfc8df7c82cacb0d8cf6fac1c9d5",
            "outputValue": 0.8321268558502197,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0002/weight.npy",
              "sha256": "1e729a87ce44d85e74fc93e6025020c65ce9266add25130f460f010f04dfbb13"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 10.312721252441406,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/output.npy",
              "sha256": "f75dfed7b35b05253cf3e7b02b9fef5490b3aad8c16bb676cfa961da6c7b3bcf"
            },
            "outputSha256": "f75dfed7b35b05253cf3e7b02b9fef5490b3aad8c16bb676cfa961da6c7b3bcf",
            "outputValue": 10.31272029876709,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0003/weight.npy",
              "sha256": "7953bf1afb06651626b3881fce2a995cae992e823ab2065e1519c61082b370e7"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.712238788604736,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/output.npy",
              "sha256": "934cf1e892927ddc90c15072f609c671b5602d9cf7e63bc233aa0b705b04f7c8"
            },
            "outputSha256": "934cf1e892927ddc90c15072f609c671b5602d9cf7e63bc233aa0b705b04f7c8",
            "outputValue": 4.7122392654418945,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0004/weight.npy",
              "sha256": "30f985ab463ba9d4a4a6f12c7a31420c1acdf93163df28e4c08552725b5d0710"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 2.879969596862793,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/output.npy",
              "sha256": "0b365c976cd6a1436d6b008d98ec35c8b249d9d0797c1a7b17c531d9a88b9548"
            },
            "outputSha256": "0b365c976cd6a1436d6b008d98ec35c8b249d9d0797c1a7b17c531d9a88b9548",
            "outputValue": 2.879969596862793,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-71629/chunk-0005/weight.npy",
              "sha256": "68d048d13f0ed17309274fe5a883db2d408e1ee43ea9b312cd83784bc6770919"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.956762313842773,
        "cpuSoftcappedLogit": 19.324335907179716,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.349609375,
        "lmHeadRowSha256": "36c405bfbe567e3e359feb0648bd30082d62f75a5b1823401b6a606168d64aef",
        "logitAbsDiff": 0.025272351872942522,
        "rawLogit": 22.956764221191406,
        "referenceRank": 26,
        "softcappedLogit": 19.324337023127057,
        "tokenId": 71629
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.340035915374756,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/output.npy",
              "sha256": "eeb038bb2236b7780ab44ab4e27eef65f3fcdadf3894cce39b7ecef16d537106"
            },
            "outputSha256": "eeb038bb2236b7780ab44ab4e27eef65f3fcdadf3894cce39b7ecef16d537106",
            "outputValue": 6.340036869049072,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0000/weight.npy",
              "sha256": "9318772a16ed255d950426dcce8f65237c8e454018fd3ea0d802ed774c76fb22"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.170402526855469,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/output.npy",
              "sha256": "64d0705cceed5fd22232b458ef3f94ed1b5c2e17c586953b80cabf9c8c5c7965"
            },
            "outputSha256": "64d0705cceed5fd22232b458ef3f94ed1b5c2e17c586953b80cabf9c8c5c7965",
            "outputValue": 6.170400142669678,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0001/weight.npy",
              "sha256": "bad3efa975717608adaeda534c2d97808d58a31ccef9133e122134a999b868b2"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.2112507820129395,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/output.npy",
              "sha256": "415dcafbc2d46693a07be6d0e87c408a6ef103c769125cd4c46182dfa0f0954f"
            },
            "outputSha256": "415dcafbc2d46693a07be6d0e87c408a6ef103c769125cd4c46182dfa0f0954f",
            "outputValue": 7.211251735687256,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0002/weight.npy",
              "sha256": "84ba193079b65fa498036cdf8c31561df64461df15fb220b2c5bf4f49761c1ac"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 16.785091400146484,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/output.npy",
              "sha256": "10182f7f0948d54582460136f39085ae371a013b11e7b9a2c35153e8795015fa"
            },
            "outputSha256": "10182f7f0948d54582460136f39085ae371a013b11e7b9a2c35153e8795015fa",
            "outputValue": 16.78508758544922,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0003/weight.npy",
              "sha256": "2330dc6bc1c4a04854315a1cbb71708243d8b278d38c5fff664c30c98a5d4081"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -13.569069862365723,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/output.npy",
              "sha256": "40a15b8996db83859945f476b17cd3e52e920088eff2e4cfa45f16f64fef79f8"
            },
            "outputSha256": "40a15b8996db83859945f476b17cd3e52e920088eff2e4cfa45f16f64fef79f8",
            "outputValue": -13.569069862365723,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0004/weight.npy",
              "sha256": "a1bb7bd2e896e5fe1d44fe2689dd1e1ef5d02dc90a8719b3d2cc1a7c1fa5a8c6"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.000209808349609375,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/output.npy",
              "sha256": "e381675d99fab32c9db0d52b9002b3198391f7952cf6454f240ef48d1a08278b"
            },
            "outputSha256": "e381675d99fab32c9db0d52b9002b3198391f7952cf6454f240ef48d1a08278b",
            "outputValue": 0.0002098381519317627,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-568/chunk-0005/weight.npy",
              "sha256": "2ba099220b367da56b9984c24d9fea6b9af8287b19926b1825cdbc1fe1828cd0"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.93791961669922,
        "cpuSoftcappedLogit": 19.313307002302096,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.294750213623047,
        "lmHeadRowSha256": "8878593d1aa1e08f8ec27f665816b5af404d52f81e9373c9a48f48f870abfabf",
        "logitAbsDiff": 0.01855455497791425,
        "rawLogit": 22.937915802001953,
        "referenceRank": 27,
        "softcappedLogit": 19.31330476860096,
        "tokenId": 568
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.383318901062012,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/output.npy",
              "sha256": "ea02ad6c2d6e70ea02b4ec294514fe989d82a5449322014bac6f183021d82ab1"
            },
            "outputSha256": "ea02ad6c2d6e70ea02b4ec294514fe989d82a5449322014bac6f183021d82ab1",
            "outputValue": 6.383318901062012,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0000/weight.npy",
              "sha256": "af0108dddd18d864cd18493f5d52eb787acb483df61610ca7e519c6f3fe9bcf3"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.889815807342529,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/output.npy",
              "sha256": "e15b367c81f36a33c624879359358c04b1abf2e64182c7fcd2ba83a36b6364c0"
            },
            "outputSha256": "e15b367c81f36a33c624879359358c04b1abf2e64182c7fcd2ba83a36b6364c0",
            "outputValue": 4.889815330505371,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0001/weight.npy",
              "sha256": "86d52a80c1b78b8e5c612ba5b5ac01246340f95ce3803efc26dcceecb05d7e8f"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 2.2616124153137207,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/output.npy",
              "sha256": "c8dbe1fc5d7377f1febeb679df092b594e332d68bc1d63ef688134972e40dab9"
            },
            "outputSha256": "c8dbe1fc5d7377f1febeb679df092b594e332d68bc1d63ef688134972e40dab9",
            "outputValue": 2.261612892150879,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0002/weight.npy",
              "sha256": "e1c3f63a192508745a36e14b422392ddbd6088a628cb6963b83a97289e67f303"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 6.540523529052734,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/output.npy",
              "sha256": "96b994f243928c830ec31b2a24aa7e834a8b7a09edd789123973d4b5a431a9cf"
            },
            "outputSha256": "96b994f243928c830ec31b2a24aa7e834a8b7a09edd789123973d4b5a431a9cf",
            "outputValue": 6.540524482727051,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0003/weight.npy",
              "sha256": "4eef1551f3ea9061dd8dc4e40e89040597dbb8528696d55b4c5b45c5360b3935"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.655866026878357,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/output.npy",
              "sha256": "a8865d44b204f62988bdaaf1b61f357ee3e4dfcf266135e15e2b1e44e871ab08"
            },
            "outputSha256": "a8865d44b204f62988bdaaf1b61f357ee3e4dfcf266135e15e2b1e44e871ab08",
            "outputValue": 1.6558665037155151,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0004/weight.npy",
              "sha256": "13adc0be20ce98dc2bb6823caf98380ade9c4069b5f2ccd2354a21505fe13e9b"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.9956496953964233,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/output.npy",
              "sha256": "c587d927f429732ff1538643ea6cd1d844379e1de906ea8ba2a9f27d91b396c7"
            },
            "outputSha256": "c587d927f429732ff1538643ea6cd1d844379e1de906ea8ba2a9f27d91b396c7",
            "outputValue": 0.9956498146057129,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2721/chunk-0005/weight.npy",
              "sha256": "a06f00739f429dc6d14c03966c917960657c08c3d91dcee00f624c2c12b692b2"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.726787567138672,
        "cpuSoftcappedLogit": 19.18911773776118,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.20282554626465,
        "lmHeadRowSha256": "e71094e7af210d9feb224fab6f7d80e0dbe7ea5199495b9f4b38cd723525b4c2",
        "logitAbsDiff": 0.013707808503468755,
        "rawLogit": 22.726787567138672,
        "referenceRank": 28,
        "softcappedLogit": 19.18911773776118,
        "tokenId": 2721
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -1.549971103668213,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/output.npy",
              "sha256": "0cc5a1c73c04c141125e3751a7378250736da03dec8c87752d1cac79116099c1"
            },
            "outputSha256": "0cc5a1c73c04c141125e3751a7378250736da03dec8c87752d1cac79116099c1",
            "outputValue": -1.5499706268310547,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0000/weight.npy",
              "sha256": "0715aca300e7136e26d626faf1fc5a65e88477b58b1aaf34ded0474d69f9323c"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.437976837158203,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/output.npy",
              "sha256": "475260f82a113cbaa66383ca63511919458ab7925d16adf67a88227184be0d77"
            },
            "outputSha256": "475260f82a113cbaa66383ca63511919458ab7925d16adf67a88227184be0d77",
            "outputValue": 8.43797492980957,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0001/weight.npy",
              "sha256": "9ec03d85b563b089352eae867f6c0dc44e0283b2b97811edb17ac9e6986f8203"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": -1.2038527727127075,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/output.npy",
              "sha256": "8f234f56a8a2c2e51a3ef2bc888b77130a5cf7811f5aae1848cf099f60588761"
            },
            "outputSha256": "8f234f56a8a2c2e51a3ef2bc888b77130a5cf7811f5aae1848cf099f60588761",
            "outputValue": -1.203852891921997,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0002/weight.npy",
              "sha256": "9f5b461b416f44fed3c4c83f2e210c68287cb1f7554312d9d396778f08128599"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 10.870891571044922,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/output.npy",
              "sha256": "29592e0c21b6cd5704cb5365cdd6069cbd740e8434de8896fd60351d81c0a49e"
            },
            "outputSha256": "29592e0c21b6cd5704cb5365cdd6069cbd740e8434de8896fd60351d81c0a49e",
            "outputValue": 10.870891571044922,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0003/weight.npy",
              "sha256": "2c627cacaa06725c6690e602bfb1df0ab94368d6ea003b2b8e9335fa4b00615e"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.815917491912842,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/output.npy",
              "sha256": "2b227628a834a044f0afa1cf12529fbff7258ddf239453b63712929ddf99cc14"
            },
            "outputSha256": "2b227628a834a044f0afa1cf12529fbff7258ddf239453b63712929ddf99cc14",
            "outputValue": 3.815917491912842,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0004/weight.npy",
              "sha256": "cba70bb94b6704c5f24af73f275e423230328e0d76fc337d9568cd1c989248a0"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 2.2962164878845215,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/output.npy",
              "sha256": "ecccff1512f5d67bc364150e500707be3b185e9a0d5b9099f02801c922875b19"
            },
            "outputSha256": "ecccff1512f5d67bc364150e500707be3b185e9a0d5b9099f02801c922875b19",
            "outputValue": 2.2962164878845215,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-7217/chunk-0005/weight.npy",
              "sha256": "77654bb4866727743655ab22f7da9647eb0f983f96fa56b57b66c226c335488e"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.667179107666016,
        "cpuSoftcappedLogit": 19.15385246117488,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.20282554626465,
        "lmHeadRowSha256": "fe3fb037d369c6d2f9f06d1bc364a9a775b591b488a691206b1bb1d5a5dc3e19",
        "logitAbsDiff": 0.04897421493943099,
        "rawLogit": 22.667177200317383,
        "referenceRank": 29,
        "softcappedLogit": 19.153851331325217,
        "tokenId": 7217
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.898571491241455,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/output.npy",
              "sha256": "03f5fd0a4d6c98845b7bf4511174d1fc07676582f3e66a385079bbab9af8aabc"
            },
            "outputSha256": "03f5fd0a4d6c98845b7bf4511174d1fc07676582f3e66a385079bbab9af8aabc",
            "outputValue": 5.898571014404297,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0000/weight.npy",
              "sha256": "aabef6a9d8c856c4f044e86ca1bf5558fdcda7da8434f7aa587440fd17d6515e"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.183839797973633,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/output.npy",
              "sha256": "c7a54dd1e5d6a55b8ca0fe159ffe03d69d15841d596448928d2b5f1b15bff563"
            },
            "outputSha256": "c7a54dd1e5d6a55b8ca0fe159ffe03d69d15841d596448928d2b5f1b15bff563",
            "outputValue": 3.183840751647949,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0001/weight.npy",
              "sha256": "8a8754be940cecee9a4d5ec1f73a19b588f2506f75af57b8f16036c9749b7cb3"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.8867753148078918,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/output.npy",
              "sha256": "95a6c65a1525c206c1d6ebc279781fdb62d680a50454638c1ae54202b94691ca"
            },
            "outputSha256": "95a6c65a1525c206c1d6ebc279781fdb62d680a50454638c1ae54202b94691ca",
            "outputValue": 0.8867745995521545,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0002/weight.npy",
              "sha256": "0ba4551034a0acc68ba89999a95106b9f6df4e2036d115921d3bca0d4a1b8827"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.0213141441345215,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/output.npy",
              "sha256": "8fff2d5daaaf394a2fdc10f1ef38b753da35e7eaee5b77499587e662c328f6d0"
            },
            "outputSha256": "8fff2d5daaaf394a2fdc10f1ef38b753da35e7eaee5b77499587e662c328f6d0",
            "outputValue": 7.021313667297363,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0003/weight.npy",
              "sha256": "d4ad161eac04202791eca253eeb23fe63d55cd23dd4e403fc2c57e8878976cec"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.8268513679504395,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/output.npy",
              "sha256": "96586a55d99b1dd376f7a3b059c486422c8de74a0286b431b1ac0cda6da41e00"
            },
            "outputSha256": "96586a55d99b1dd376f7a3b059c486422c8de74a0286b431b1ac0cda6da41e00",
            "outputValue": 5.826850414276123,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0004/weight.npy",
              "sha256": "a7b162d18f53eebf6ad2d6abeb0b7cfe29039840b3b37e4b745ec999bbd9fd9a"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.19219154119491577,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/output.npy",
              "sha256": "13611f2839dd59640f9145a44feda02d7434eef64f75e056c7adeb0502c7a4c9"
            },
            "outputSha256": "13611f2839dd59640f9145a44feda02d7434eef64f75e056c7adeb0502c7a4c9",
            "outputValue": -0.19219163060188293,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2462/chunk-0005/weight.npy",
              "sha256": "b79745d5f22ec4e58155e03b2e8d14a640756b69f230e55ef5d5e7e53e67421f"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.625160217285156,
        "cpuSoftcappedLogit": 19.12893961214963,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.110286712646484,
        "lmHeadRowSha256": "49d4aab345cad1d586c43851ed33ba77447d8ea84e2e13be4ac17d71d1d7d563",
        "logitAbsDiff": 0.018652899503145193,
        "rawLogit": 22.625160217285156,
        "referenceRank": 30,
        "softcappedLogit": 19.12893961214963,
        "tokenId": 2462
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.9335832595825195,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/output.npy",
              "sha256": "7c781e15eea42cb5f8fd724aa04e7d000fee4dbd395eb5501e6d242e5846667f"
            },
            "outputSha256": "7c781e15eea42cb5f8fd724aa04e7d000fee4dbd395eb5501e6d242e5846667f",
            "outputValue": 4.933585166931152,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0000/weight.npy",
              "sha256": "202b142f0aac00670ec80b79ba3c0be38ead6fccf34992eb186085edca1899e2"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 3.438114881515503,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/output.npy",
              "sha256": "cf608d3aad3997b6d2942019a6b99c02dd9ba52467f5cbe0ed35041a32ebfac8"
            },
            "outputSha256": "cf608d3aad3997b6d2942019a6b99c02dd9ba52467f5cbe0ed35041a32ebfac8",
            "outputValue": 3.438114643096924,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0001/weight.npy",
              "sha256": "d375ed85dd03cd1384506e522c2f0b38162995909bd441a07e2eb65d632c3ac3"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.609157562255859,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/output.npy",
              "sha256": "5919d4f4ef48b19e5fe2f6afb7aeb8c997d3fba82e92ab470e4216aa1eaa4e5d"
            },
            "outputSha256": "5919d4f4ef48b19e5fe2f6afb7aeb8c997d3fba82e92ab470e4216aa1eaa4e5d",
            "outputValue": 5.609156608581543,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0002/weight.npy",
              "sha256": "66e543d522aac41c84ecf7f7af4ed1e27fe99627c9fcca1edd0d03dfde0c7b07"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.179441452026367,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/output.npy",
              "sha256": "d262f66ffaf158a34238a15fe46724a5c3fc393449991e6a43e9ecc56b33eb69"
            },
            "outputSha256": "d262f66ffaf158a34238a15fe46724a5c3fc393449991e6a43e9ecc56b33eb69",
            "outputValue": 4.179442405700684,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0003/weight.npy",
              "sha256": "128d545e608ae42faa349da6520b8ad4ad7e8d3b0909875fac1a66e838951af1"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 4.238471984863281,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/output.npy",
              "sha256": "f3269654cdbb6884b2510a62bb2ff694dfe300199e5f4240c5e62758f571d9d1"
            },
            "outputSha256": "f3269654cdbb6884b2510a62bb2ff694dfe300199e5f4240c5e62758f571d9d1",
            "outputValue": 4.238472938537598,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0004/weight.npy",
              "sha256": "50f8f56be7ac21ccc4c418783cd2d237014ddf71445772e6f1a8741a27973369"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": 0.13431495428085327,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/output.npy",
              "sha256": "4762e28e3fee74d0f35ce4dcf8c7f2b2a2e8d034e6cab5650f1fb38f3c418d6c"
            },
            "outputSha256": "4762e28e3fee74d0f35ce4dcf8c7f2b2a2e8d034e6cab5650f1fb38f3c418d6c",
            "outputValue": 0.1343151032924652,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-14769/chunk-0005/weight.npy",
              "sha256": "05c413f5db1f5b7f1c2ed2e1e45f2816ef8b81c880f29b2d511b3c8c12e5e6c9"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.533084869384766,
        "cpuSoftcappedLogit": 19.074192711182008,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.08240509033203,
        "lmHeadRowSha256": "7f76a8a60505a40ed6984dfe3ac7b5dd6cf7eb6f1199986456e9b2a2d54f0d8c",
        "logitAbsDiff": 0.008210106543586448,
        "rawLogit": 22.53308868408203,
        "referenceRank": 31,
        "softcappedLogit": 19.074194983788445,
        "tokenId": 14769
      },
      {
        "allChunksSucceeded": true,
        "chunkCount": 6,
        "chunkResults": [
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/activation.npy",
              "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
            },
            "chunkIndex": 0,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 5.275421142578125,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 0,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/output.npy",
              "sha256": "04688e7267b98a84873483c51ae2c73bb9a71ec57e2a68c2310e6f7892e509de"
            },
            "outputSha256": "04688e7267b98a84873483c51ae2c73bb9a71ec57e2a68c2310e6f7892e509de",
            "outputValue": 5.275421619415283,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0000/weight.npy",
              "sha256": "5998515592fb2e4ffca3d62cb0c0fa7cd03b2d553be94ed84d7ad36d4f4a33df"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/activation.npy",
              "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
            },
            "chunkIndex": 1,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 8.22639274597168,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 1024,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/output.npy",
              "sha256": "ee89aba718973740cd1fd10ffcb57325844339454058fae97bc27bab034e6041"
            },
            "outputSha256": "ee89aba718973740cd1fd10ffcb57325844339454058fae97bc27bab034e6041",
            "outputValue": 8.226393699645996,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0001/weight.npy",
              "sha256": "60c838b7dcd98589062b163bc19325b64078e8bdba9cc9b8b2baf4263fba8562"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/activation.npy",
              "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
            },
            "chunkIndex": 2,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 1.3599804639816284,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 2048,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/output.npy",
              "sha256": "647ece2e99859888e1441f1bb24f8ce671285dbfc10ca8e769d9e20b464cf41c"
            },
            "outputSha256": "647ece2e99859888e1441f1bb24f8ce671285dbfc10ca8e769d9e20b464cf41c",
            "outputValue": 1.359980821609497,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0002/weight.npy",
              "sha256": "ff3e5aafa93dd15fede3a52e4d9a05088d99468eefab0e82a44228c6e13047ef"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/activation.npy",
              "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
            },
            "chunkIndex": 3,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 7.280179977416992,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 3072,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/output.npy",
              "sha256": "214ad1aa6c7d0452700d6528ce2587d05308527fa500c4648087e0201523a441"
            },
            "outputSha256": "214ad1aa6c7d0452700d6528ce2587d05308527fa500c4648087e0201523a441",
            "outputValue": 7.280179500579834,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0003/weight.npy",
              "sha256": "a35b77606f06a5f37ec879b2f7c90ed2c6f08e3950f8cba43f7f028efe9355c0"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/activation.npy",
              "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
            },
            "chunkIndex": 4,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
              "--width",
              "32",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/output.npy:f32:1:31,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
            "cpuPartial": 0.6420308351516724,
            "exitCode": 0,
            "hiddenCount": 1024,
            "hiddenStart": 4096,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/output.npy",
              "sha256": "fec2b0eea3ed68bced2e072728abef8b2a7dc6b79e0d2c183cc5417776e8b92a"
            },
            "outputSha256": "fec2b0eea3ed68bced2e072728abef8b2a7dc6b79e0d2c183cc5417776e8b92a",
            "outputValue": 0.642030656337738,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 711,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/phase.log",
              "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 2176,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0004/weight.npy",
              "sha256": "9e85b7ce3fe778f8accfba1959871f33aeaac3a30923fe10485eb79630d1cd12"
            },
            "width": 32
          },
          {
            "activation": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/activation.npy",
              "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
            },
            "chunkIndex": 5,
            "command": [
              "<cerebras-sdk-root>/cs_python",
              "bench/runners/csl-runners/chain_step_adapter.py",
              "--compile-dir",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
              "--width",
              "8",
              "--height",
              "1",
              "--chunk-size",
              "32",
              "--input",
              "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/activation.npy:f16:32",
              "--input",
              "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/weight.npy:f16:32",
              "--output",
              "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/output.npy:f32:1:7,0,1,1",
              "--phase-trace",
              "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/phase.log"
            ],
            "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
            "cpuPartial": -0.21671369671821594,
            "exitCode": 0,
            "hiddenCount": 256,
            "hiddenStart": 5120,
            "output": {
              "byteLength": 132,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/output.npy",
              "sha256": "77ea91c55a3829b26d30e72337eea2f77c0ddf0b9e5120888243372dd25cc60a"
            },
            "outputSha256": "77ea91c55a3829b26d30e72337eea2f77c0ddf0b9e5120888243372dd25cc60a",
            "outputValue": -0.216713547706604,
            "phaseTail": [
              "phase:step_start step=0",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=activation",
              "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
              "phase:memcpy_h2d_complete step=0 symbol=weight",
              "phase:launch_start function=compute step=0",
              "phase:launch_complete function=compute step=0",
              "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
              "phase:memcpy_d2h_complete step=0 symbol=output",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "phaseTrace": {
              "byteLength": 708,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/phase.log",
              "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
            },
            "status": "succeeded",
            "stderrTail": [
              "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
              "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
            ],
            "stdoutTail": [
              "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/output.npy (float32 shape=(1,))",
              "phase:step_complete step=0",
              "phase:stop_start",
              "phase:stop_complete"
            ],
            "timedOut": false,
            "weight": {
              "byteLength": 640,
              "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-2173/chunk-0005/weight.npy",
              "sha256": "f955b9206255ed7afaf4295462761416b1dee247976f7373a669448e0d562c9b"
            },
            "width": 8
          }
        ],
        "cpuRawLogit": 22.567291259765625,
        "cpuSoftcappedLogit": 19.09455640322418,
        "cslCpuPartialsMatch": true,
        "expectedLogit": 19.07309913635254,
        "lmHeadRowSha256": "39a8cdd47218c8fe54bd3e3e3d95f2643dc924530c2527aca6233773e33aa981",
        "logitAbsDiff": 0.02145840152768841,
        "rawLogit": 22.567293167114258,
        "referenceRank": 32,
        "softcappedLogit": 19.094557537880227,
        "tokenId": 2173
      }
    ],
    "candidateTokenIds": [
      3730,
      4781,
      5213,
      11082,
      496,
      3187,
      2214,
      236858,
      14102,
      6816,
      1390,
      506,
      9595,
      1144,
      9199,
      614,
      37155,
      3643,
      5633,
      1607,
      7205,
      236743,
      236789,
      236772,
      1401,
      71629,
      568,
      2721,
      7217,
      2462,
      14769,
      2173
    ],
    "chunkCount": 6,
    "chunkPeWidth": 32,
    "chunkResults": [
      {
        "activation": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/activation.npy",
          "sha256": "c66e33d94e44a6a8c8aa49b1458d2b513ef7e8eb1ffca47ddbeed6dbb28cddca"
        },
        "chunkIndex": 0,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
          "--width",
          "32",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy:f32:1:31,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
        "cpuPartial": 13.98351001739502,
        "exitCode": 0,
        "hiddenCount": 1024,
        "hiddenStart": 0,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy",
          "sha256": "340351e8d251f0d633227deaa299ac93c0ea5cdbc05260087b89471a7f44c63a"
        },
        "outputSha256": "340351e8d251f0d633227deaa299ac93c0ea5cdbc05260087b89471a7f44c63a",
        "outputValue": 13.983509063720703,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 711,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/phase.log",
          "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0000/weight.npy",
          "sha256": "2d23d186a027a06ae5b3ca5076844bf22d646145b6c66a65ecb55ed855f8a0bd"
        },
        "width": 32
      },
      {
        "activation": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/activation.npy",
          "sha256": "2a9e0066d4fd89015f03f99cb026f6347dc807e27dd6a49cebee7b4faa7a081b"
        },
        "chunkIndex": 1,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
          "--width",
          "32",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy:f32:1:31,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
        "cpuPartial": 15.017446517944336,
        "exitCode": 0,
        "hiddenCount": 1024,
        "hiddenStart": 1024,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy",
          "sha256": "83530e337324211aad835854eddc482634f71200ab6a1bfdaf0e32c40abac15d"
        },
        "outputSha256": "83530e337324211aad835854eddc482634f71200ab6a1bfdaf0e32c40abac15d",
        "outputValue": 15.017444610595703,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 711,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/phase.log",
          "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0001/weight.npy",
          "sha256": "5623016a9e6cc8774e996dba7e617c6c29421f283f2b501c2926a125a59ad2c7"
        },
        "width": 32
      },
      {
        "activation": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/activation.npy",
          "sha256": "9ea40b530a4cfe379768f5f9dfa39f31250644cf12e27ec4fe4d2611f941ce29"
        },
        "chunkIndex": 2,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
          "--width",
          "32",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy:f32:1:31,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
        "cpuPartial": 7.004585266113281,
        "exitCode": 0,
        "hiddenCount": 1024,
        "hiddenStart": 2048,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy",
          "sha256": "3829aaf82a1ff1e1727a03186da677c943c4539935b52532457d6161b6f6b5fc"
        },
        "outputSha256": "3829aaf82a1ff1e1727a03186da677c943c4539935b52532457d6161b6f6b5fc",
        "outputValue": 7.004586219787598,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 711,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/phase.log",
          "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0002/weight.npy",
          "sha256": "9682b6c538b0ab1eeefdc084c15ea71e72ae06903c9c9e573012b0db1a738011"
        },
        "width": 32
      },
      {
        "activation": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/activation.npy",
          "sha256": "4149b0074b8744ddb30acea597868ab2750d99d22e00c10a2b99db9df4b1916e"
        },
        "chunkIndex": 3,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
          "--width",
          "32",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy:f32:1:31,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
        "cpuPartial": 13.789728164672852,
        "exitCode": 0,
        "hiddenCount": 1024,
        "hiddenStart": 3072,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy",
          "sha256": "f8375119663126093a1790b0d1eeb72bf8a1c35ff7080898f182660001593b6e"
        },
        "outputSha256": "f8375119663126093a1790b0d1eeb72bf8a1c35ff7080898f182660001593b6e",
        "outputValue": 13.789728164672852,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 711,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/phase.log",
          "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0003/weight.npy",
          "sha256": "695a88ac3ee08948f1f00de41984183087523f79330221311f241408b51fa8eb"
        },
        "width": 32
      },
      {
        "activation": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/activation.npy",
          "sha256": "3f8c1baa9705fa28c4406a40388443ae9bafb10cf433dca062f25bac3d8bf48a"
        },
        "chunkIndex": 4,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
          "--width",
          "32",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy:f32:1:31,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0032/compiled",
        "cpuPartial": 4.878942489624023,
        "exitCode": 0,
        "hiddenCount": 1024,
        "hiddenStart": 4096,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy",
          "sha256": "06428400630673a00f1fc6316780d30e56e93773761d338988881cf203688fd9"
        },
        "outputSha256": "06428400630673a00f1fc6316780d30e56e93773761d338988881cf203688fd9",
        "outputValue": 4.87894344329834,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=32 words=512 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=31 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 711,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/phase.log",
          "sha256": "1f384e0a0def8fa4db7dc9da801452f5b1e15efc1393ca8182ba75fee99b9575"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 2176,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0004/weight.npy",
          "sha256": "daa29d2e6e0fca7c94f9169594725b04d9f5580160c83e6ff28930ccc44dca98"
        },
        "width": 32
      },
      {
        "activation": {
          "byteLength": 640,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/activation.npy",
          "sha256": "f4ea6f28ed7aaccb584cf7a1ec0b7bf3e409e44510050a686e8b77f51346d159"
        },
        "chunkIndex": 5,
        "command": [
          "<cerebras-sdk-root>/cs_python",
          "bench/runners/csl-runners/chain_step_adapter.py",
          "--compile-dir",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
          "--width",
          "8",
          "--height",
          "1",
          "--chunk-size",
          "32",
          "--input",
          "activation:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/activation.npy:f16:32",
          "--input",
          "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/weight.npy:f16:32",
          "--output",
          "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy:f32:1:7,0,1,1",
          "--phase-trace",
          "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/phase.log"
        ],
        "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-w0008/compiled",
        "cpuPartial": 2.9069056510925293,
        "exitCode": 0,
        "hiddenCount": 256,
        "hiddenStart": 5120,
        "output": {
          "byteLength": 132,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy",
          "sha256": "a09ee00c22b4474bbfb9045c987eb39976f2a4b2df5d06e18412a683d11df99c"
        },
        "outputSha256": "a09ee00c22b4474bbfb9045c987eb39976f2a4b2df5d06e18412a683d11df99c",
        "outputValue": 2.9069058895111084,
        "phaseTail": [
          "phase:step_start step=0",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=activation width=8 words=128 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=activation",
          "phase:memcpy_h2d_start chunk=16 height=1 step=0 symbol=weight width=8 words=128 x=0 y=0",
          "phase:memcpy_h2d_complete step=0 symbol=weight",
          "phase:launch_start function=compute step=0",
          "phase:launch_complete function=compute step=0",
          "phase:memcpy_d2h_start chunk=1 height=1 step=0 symbol=output width=1 words=1 x=7 y=0",
          "phase:memcpy_d2h_complete step=0 symbol=output",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "phaseTrace": {
          "byteLength": 708,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/phase.log",
          "sha256": "d5f7fbf4ec9413993ca7eeaef4bb6bd787b2e85682a7326d665e70e7f436bf80"
        },
        "status": "succeeded",
        "stderrTail": [
          "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
          "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
        ],
        "stdoutTail": [
          "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/output.npy (float32 shape=(1,))",
          "phase:step_complete step=0",
          "phase:stop_start",
          "phase:stop_complete"
        ],
        "timedOut": false,
        "weight": {
          "byteLength": 640,
          "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/token-3730/chunk-0005/weight.npy",
          "sha256": "bf8782b5ebda2d0b0f7c6154467fa4d27bb6b9de22d41fb67e448f0ecf0a651c"
        },
        "width": 8
      }
    ],
    "cpuRawLogit": 57.58111572265625,
    "cpuSoftcappedLogit": 28.735962280829014,
    "cslArgmaxTokenId": 3730,
    "decisionMarginLowerBound": 3.6867826176480207,
    "finalNorm": {
      "atol": 0.005,
      "command": [
        "<cerebras-sdk-root>/cs_python",
        "bench/runners/csl-runners/chain_step_adapter.py",
        "--compile-dir",
        "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-final-norm-f16/compiled",
        "--width",
        "1",
        "--height",
        "1",
        "--chunk-size",
        "5376",
        "--input",
        "input:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/input.npy:f16:5376",
        "--input",
        "weight:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/weight.npy:f16:5376",
        "--output",
        "output:bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/output.npy:f16:5376:0,0,1,1",
        "--phase-trace",
        "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/phase.log"
      ],
      "compileDir": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/compile-final-norm-f16/compiled",
      "exactMatchVsHostF16": false,
      "exitCode": 0,
      "hiddenSize": 5376,
      "input": {
        "byteLength": 10880,
        "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/input.npy",
        "sha256": "f4232af112459da6c8ddca30aafa011fffd6619a3888349163145b8075a76990"
      },
      "kernel": "final_norm_f16",
      "maxAbsDiffVsHostF16": 0.00390625,
      "output": {
        "byteLength": 10880,
        "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/output.npy",
        "sha256": "aae428fbb005106c10191575f0cb193d77b0e1d8d501d1802b24a6ca229ff5ef"
      },
      "outputSha256": "aae428fbb005106c10191575f0cb193d77b0e1d8d501d1802b24a6ca229ff5ef",
      "phaseTail": [
        "phase:step_start step=0",
        "phase:memcpy_h2d_start chunk=2688 height=1 step=0 symbol=input width=1 words=2688 x=0 y=0",
        "phase:memcpy_h2d_complete step=0 symbol=input",
        "phase:memcpy_h2d_start chunk=2688 height=1 step=0 symbol=weight width=1 words=2688 x=0 y=0",
        "phase:memcpy_h2d_complete step=0 symbol=weight",
        "phase:launch_start function=compute step=0",
        "phase:launch_complete function=compute step=0",
        "phase:memcpy_d2h_start chunk=2688 height=1 step=0 symbol=output width=1 words=2688 x=0 y=0",
        "phase:memcpy_d2h_complete step=0 symbol=output",
        "phase:step_complete step=0",
        "phase:stop_start",
        "phase:stop_complete"
      ],
      "phaseTrace": {
        "byteLength": 710,
        "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/phase.log",
        "sha256": "8e2e523a6a0aa82098057208eb770cab2d30593c154de684e474ade9632559a2"
      },
      "status": "succeeded",
      "stderrTail": [
        "\u001b[34m[INFO]\u001b[0m === Calling container-hosted python ===",
        "\u001b[34m[INFO]\u001b[0m Using direct rootfs: <cerebras-sdk-root>/.direct-rootfs"
      ],
      "stdoutTail": [
        "[adapter] wrote output -> bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/output.npy (float16 shape=(5376,))",
        "phase:step_complete step=0",
        "phase:stop_start",
        "phase:stop_complete"
      ],
      "timedOut": false,
      "weight": {
        "byteLength": 10880,
        "path": "bench/out/r3-1-31b-af16-doppler-csl-splice/selected-logit-splice/final-norm-csl/weight.npy",
        "sha256": "3242b20a07e2ffa1227e5a5c930c745d05fb2c954827c25a49dec113ebe67212"
      },
      "withinAtol": true
    },
    "inDimPerPe": 32,
    "kernel": "lm_head_prefill",
    "logitAbsDiff": 0.008746734063564077,
    "maxLogitAbsDiff": 0.11128245826583338,
    "rawLogit": 57.58111572265625,
    "referenceArgmaxTokenId": 3730,
    "referenceTop1Top2Margin": 3.9093475341796875,
    "rmsNormEps": 1e-06,
    "softcap": 30.0,
    "softcappedLogit": 28.735962280829014,
    "strictLogitToleranceAtol": 0.02,
    "strictLogitTolerancePassed": false,
    "tailKernels": [
      "final_norm_f16",
      "lm_head_prefill"
    ],
    "topK": 32
  },
  "dopplerReference": {
    "expectedSelectedLogit": 28.744709014892578,
    "expectedTokenId": 3730,
    "expectedTopKLogits": [
      {
        "logit": 28.744709014892578,
        "rank": 1,
        "tokenId": 3730
      },
      {
        "logit": 24.83536148071289,
        "rank": 2,
        "tokenId": 4781
      },
      {
        "logit": 24.18125343322754,
        "rank": 3,
        "tokenId": 5213
      },
      {
        "logit": 24.037384033203125,
        "rank": 4,
        "tokenId": 11082
      },
      {
        "logit": 23.763450622558594,
        "rank": 5,
        "tokenId": 496
      },
      {
        "logit": 22.958629608154297,
        "rank": 6,
        "tokenId": 3187
      },
      {
        "logit": 22.94567108154297,
        "rank": 7,
        "tokenId": 2214
      },
      {
        "logit": 22.403263092041016,
        "rank": 8,
        "tokenId": 236858
      },
      {
        "logit": 21.594894409179688,
        "rank": 9,
        "tokenId": 14102
      },
      {
        "logit": 21.101390838623047,
        "rank": 10,
        "tokenId": 6816
      },
      {
        "logit": 20.926305770874023,
        "rank": 11,
        "tokenId": 1390
      },
      {
        "logit": 20.82151985168457,
        "rank": 12,
        "tokenId": 506
      },
      {
        "logit": 20.780956268310547,
        "rank": 13,
        "tokenId": 9595
      },
      {
        "logit": 20.756546020507812,
        "rank": 14,
        "tokenId": 1144
      },
      {
        "logit": 20.674800872802734,
        "rank": 15,
        "tokenId": 9199
      },
      {
        "logit": 20.584197998046875,
        "rank": 16,
        "tokenId": 614
      },
      {
        "logit": 20.509538650512695,
        "rank": 17,
        "tokenId": 37155
      },
      {
        "logit": 20.45113754272461,
        "rank": 18,
        "tokenId": 3643
      },
      {
        "logit": 20.367202758789062,
        "rank": 19,
        "tokenId": 5633
      },
      {
        "logit": 20.291152954101562,
        "rank": 20,
        "tokenId": 1607
      },
      {
        "logit": 19.78056526184082,
        "rank": 21,
        "tokenId": 7205
      },
      {
        "logit": 19.602706909179688,
        "rank": 22,
        "tokenId": 236743
      },
      {
        "logit": 19.458667755126953,
        "rank": 23,
        "tokenId": 236789
      },
      {
        "logit": 19.40424919128418,
        "rank": 24,
        "tokenId": 236772
      },
      {
        "logit": 19.376956939697266,
        "rank": 25,
        "tokenId": 1401
      },
      {
        "logit": 19.349609375,
        "rank": 26,
        "tokenId": 71629
      },
      {
        "logit": 19.294750213623047,
        "rank": 27,
        "tokenId": 568
      },
      {
        "logit": 19.20282554626465,
        "rank": 28,
        "tokenId": 2721
      },
      {
        "logit": 19.20282554626465,
        "rank": 29,
        "tokenId": 7217
      },
      {
        "logit": 19.110286712646484,
        "rank": 30,
        "tokenId": 2462
      },
      {
        "logit": 19.08240509033203,
        "rank": 31,
        "tokenId": 14769
      },
      {
        "logit": 19.07309913635254,
        "rank": 32,
        "tokenId": 2173
      }
    ],
    "expectedTopKTokenIds": [
      3730,
      4781,
      5213,
      11082,
      496,
      3187,
      2214,
      236858,
      14102,
      6816,
      1390,
      506,
      9595,
      1144,
      9199,
      614,
      37155,
      3643,
      5633,
      1607,
      7205,
      236743,
      236789,
      236772,
      1401,
      71629,
      568,
      2721,
      7217,
      2462,
      14769,
      2173
    ],
    "fixtureManifest": {
      "byteLength": 3735,
      "path": "bench/fixtures/r3-1-31b-doppler-frozen-af16/frozen-reference.manifest.json",
      "sha256": "8a92d64c1b388c3d663cc3b486305b2306be8c26aa9ff165b5aa73c5095f9f84"
    },
    "generatedTokenIds": [
      3730,
      236761
    ],
    "inputTensor": {
      "byteLength": 150656,
      "path": "bench/fixtures/r3-1-31b-doppler-frozen-af16/layer_59/post_ffn.npy",
      "sha256": "7c6d01ed06482b057a21dfebc8ba0e68f99e83fb50d90733971e02d0153ef19d"
    },
    "prefillLogits": {
      "byteLength": 1048576,
      "path": "bench/out/doppler-reference/gemma-4-31b-af16-bos-the-color-of-the-sky-is-prefill-decode2/final_logits.f32",
      "sha256": "f1e5c4ff139d3ce7fdd1f42d02014f33b1a11071232ea5bd8fc422e5d02319a7"
    }
  },
  "manifestPath": "<scrubbed>",
  "manifestSha256": "8d71be2650a1bd35f38111e6cf670b52a55aa80f9f22497ddb38a92c7398a9f8",
  "modelId": "gemma-4-31b-it-text-q4k-ehf16-af16",
  "receiptClass": "manifest_shape_doppler_selected_logit_splice",
  "referenceFixtureHash": "db16b94c49b74eb2701824f648ea6e48bae2d11d5d1c285ea938d18bf7fc786f",
  "schemaVersion": 1,
  "sourceProgram": {
    "authoringSurface": "doppler_execution_v1",
    "executionGraphSha256": "9c462d2437a77e181a2deb0ae81af8f8545677d1a881cac7c7fa4d3692bb0ab9",
    "inputSetSha256": "7b350257a61e31019ca621d6a27f7daac1b5f4e2ed69c9cf0be3557ce32d3e80",
    "manifestSha256": "8d71be2650a1bd35f38111e6cf670b52a55aa80f9f22497ddb38a92c7398a9f8",
    "referenceExport": {
      "byteLength": 100414,
      "path": "bench/out/doppler-reference/gemma-4-31b-af16-bos-the-color-of-the-sky-is-prefill-decode2/doppler_int4ple_reference_export.json",
      "sha256": "8778d4781d7fa06d214457abc1fb18cdbbc67e83fa9efbe2f67e6df4af73df55"
    },
    "weightSetSha256": "24f0c6a070fa93fe6f84e2457803cf36b34b3d0fe1de56031755f686928896bd"
  },
  "splicePoint": {
    "inputProbe": "post_ffn",
    "kind": "selected_lm_head_logit",
    "layerIndex": 59,
    "promptTokenCount": 7,
    "selectedText": " blue",
    "selectedTokenId": 3730,
    "selectedTokenIds": [
      3730,
      4781,
      5213,
      11082,
      496,
      3187,
      2214,
      236858,
      14102,
      6816,
      1390,
      506,
      9595,
      1144,
      9199,
      614,
      37155,
      3643,
      5633,
      1607,
      7205,
      236743,
      236789,
      236772,
      1401,
      71629,
      568,
      2721,
      7217,
      2462,
      14769,
      2173
    ],
    "topK": 32
  },
  "verdict": "pass",
  "weights": {
    "candidateLmHeadRows": [
      {
        "sha256": "2ebf2c44877f6ad0ca24484b636db369aca76bea856ae383b8dde725fb707032",
        "tokenId": 3730
      },
      {
        "sha256": "03c687533b180ada318278d472820d56419293d1afc868cfefcc7ae2b879a25a",
        "tokenId": 4781
      },
      {
        "sha256": "f82dc96af0c0da89f87ae82bb94088b338839cb9f025b64b1d4b960b5168fc04",
        "tokenId": 5213
      },
      {
        "sha256": "ad556d7bfcac36675bb71fa94622b58b27bd12e458335df8a373ba4824620cac",
        "tokenId": 11082
      },
      {
        "sha256": "c597008e60d1cd69ec0aae0d44bd620be1b90779907832c1621ea5d965c03b22",
        "tokenId": 496
      },
      {
        "sha256": "4623aa3406178cfed7556bc47582b3ce20cf29661a4494a882f4e8e05123f733",
        "tokenId": 3187
      },
      {
        "sha256": "28ded977f346580b36569f0f7418b823978df8674d105bd33e134cc3c6014879",
        "tokenId": 2214
      },
      {
        "sha256": "8a881316ace0dffd627cc559c458d5e779c93b3894ae929e1455f25e897685b2",
        "tokenId": 236858
      },
      {
        "sha256": "915ebb3abb4526046536e1b40bc528fc01574869c0a30096f6a3d043ab9a991f",
        "tokenId": 14102
      },
      {
        "sha256": "781d05d90c05db41305dc8b201a142cee76e4474019979c3b0c5eb4d19fed89d",
        "tokenId": 6816
      },
      {
        "sha256": "3a8aecb26db8c89ef37f4fcf1d1550d05f8890999336bcb9614c411c7d69392a",
        "tokenId": 1390
      },
      {
        "sha256": "1f7bf1bf2a9614e5ccd8e0742de914e0e4f8f952faad6db04d7951b4c7391899",
        "tokenId": 506
      },
      {
        "sha256": "1778f7b7306eba4b567a52e6cd368b88366bb6df85a4412f7f1bac628cce91e4",
        "tokenId": 9595
      },
      {
        "sha256": "d0e7518f58c7246a455d1dd2997463975c0dc997615fcac555cfd24967c60d57",
        "tokenId": 1144
      },
      {
        "sha256": "003d20406f1b57f873f4c82da989c53abd3d970d8f2e625d1412c9846305a688",
        "tokenId": 9199
      },
      {
        "sha256": "f6bfb92d6788249b45584dbdd3785f7cd7426e45f9a1caf95251dc37626d8844",
        "tokenId": 614
      },
      {
        "sha256": "2505112d2cf6eb980974dbee2af93a20470ab52352237800283227c66efc5996",
        "tokenId": 37155
      },
      {
        "sha256": "f39ee9a470c8c9011ac674939355c2fe4f93c49ca95c721a3e775991de2dcce4",
        "tokenId": 3643
      },
      {
        "sha256": "d469ea4e7b6248a5124935fa3d29814621740c86e19891997657bb24b5e4506a",
        "tokenId": 5633
      },
      {
        "sha256": "5f817cf5c7afaa7c69d7dbefc3cde32aa47d191cd5c15705231b332ea43fc39e",
        "tokenId": 1607
      },
      {
        "sha256": "d9bb31e2ebf1f6b472070ee42555c3a92dcfc240f6d345d40348ed8f28b989c3",
        "tokenId": 7205
      },
      {
        "sha256": "1c473edc5651775658fad0710cf6cc8294c8a0885677f3d262298c0fd3dcff59",
        "tokenId": 236743
      },
      {
        "sha256": "98d8d5f8a86440666160bc06f3e9bcdecc4d9ec2db4559cdc6a24994bc7a3519",
        "tokenId": 236789
      },
      {
        "sha256": "95442643b6799ee12d22cc765335c8d5cdee1e034e463e923cf90bdd174db621",
        "tokenId": 236772
      },
      {
        "sha256": "e2874fa188a1a771be5c2c8589c266b8b3193eda838504bfc95c8e46fdbc0d2a",
        "tokenId": 1401
      },
      {
        "sha256": "36c405bfbe567e3e359feb0648bd30082d62f75a5b1823401b6a606168d64aef",
        "tokenId": 71629
      },
      {
        "sha256": "8878593d1aa1e08f8ec27f665816b5af404d52f81e9373c9a48f48f870abfabf",
        "tokenId": 568
      },
      {
        "sha256": "e71094e7af210d9feb224fab6f7d80e0dbe7ea5199495b9f4b38cd723525b4c2",
        "tokenId": 2721
      },
      {
        "sha256": "fe3fb037d369c6d2f9f06d1bc364a9a775b591b488a691206b1bb1d5a5dc3e19",
        "tokenId": 7217
      },
      {
        "sha256": "49d4aab345cad1d586c43851ed33ba77447d8ea84e2e13be4ac17d71d1d7d563",
        "tokenId": 2462
      },
      {
        "sha256": "7f76a8a60505a40ed6984dfe3ac7b5dd6cf7eb6f1199986456e9b2a2d54f0d8c",
        "tokenId": 14769
      },
      {
        "sha256": "39a8cdd47218c8fe54bd3e3e3d95f2643dc924530c2527aca6233773e33aa981",
        "tokenId": 2173
      }
    ],
    "finalNormCslOutputSha256": "aae428fbb005106c10191575f0cb193d77b0e1d8d501d1802b24a6ca229ff5ef",
    "finalNormSha256": "02bde86fee13be1e7d898e2d0b60306002d6a1f09e64706ef575e22054517663",
    "finalNormTensor": "model.language_model.norm.weight",
    "lmHeadTensor": "model.language_model.embed_tokens.weight",
    "lmHeadTiedEmbedding": true,
    "selectedLmHeadRowSha256": "2ebf2c44877f6ad0ca24484b636db369aca76bea856ae383b8dde725fb707032",
    "sourceRoot": "<scrubbed>"
  }
}