Run 01a0202b · KernelIndex

Primary measurement

7.21msmedian

Rank 1 in its comparison group · source-native comparison · observed 2025-03-03

ms 20 7.206470489501953 · ms 50 7.213727951049805 · ms 80 7.229574680328369

Identity

implementationTransformers

projectHugging Face Transformers

revisionunknown

workloadseq = 512 · batch = 4 · vocab = 128256 · hidden = 1024 · bf16

comparison keysha256:278538dfe4ea78e8…

sourceLiger-Kernel benchmarks

external idkto_loss/forward/huggingface/nvidia-h100-80gb-hbm3/batch4-hidden1024-seq512-vocab128256-bf16/2025-03-03-08-23-01

sha256:434b44a70c4b5cf0082e93c…

Correctness

Marked passed by the source; the correctness policy was not published.

Workload

seq512

batch4

vocab128256

hidden1024

xbf16 [4, 512, 1024]

klbf16 [1]

biasbf16 [128256]

ref_xbf16 [4, 512, 1024]

targetint64 [4, 512]

weightbf16 [128256, 1024]

ref_biasbf16 [128256]

ref_weightbf16 [128256, 1024]

preference_labelsbool [4]

definition comparatornot_asserted

Measurements

latency · median7.21 ms

latency · p207.21 ms

latency · p807.23 ms

Protocol

harnessLiger-Kernel benchmark scripts

timercuda_events

primaryStatisticmedian

comparabilityFamilyliger_kernel_bench

Environment

gpuNVIDIA H100 (sm_90)

Artifacts

No artifacts published with this run.

Replications and notes

No attestations yet.

Community attestations. They never change the evidence level; only a KernelIndex-controlled rerun does.

Add a reproduction or note

Canonical manifest

Show manifest

{
  "run": {
    "kind": "BenchmarkRun",
    "spec": {
      "status": "passed",
      "timing": {
        "latencyNs": {
          "median": 7213728
        },
        "primaryStatistic": "median"
      },
      "observedAt": "2025-03-03T08:23:01.000Z",
      "measurements": [
        {
          "unit": "ns",
          "value": 7206470,
          "metric": "latency",
          "statistic": "p20"
        },
        {
          "unit": "ns",
          "value": 7229575,
          "metric": "latency",
          "statistic": "p80"
        }
      ],
      "sourceNative": {
        "source": "liger-kernel-bench",
        "metrics": {
          "ms_20": 7.206470489501953,
          "ms_50": 7.213727951049805,
          "ms_80": 7.229574680328369
        },
        "benchmark": "kto_loss/forward",
        "externalId": "kto_loss/forward/huggingface/nvidia-h100-80gb-hbm3/batch4-hidden1024-seq512-vocab128256-bf16/2025-03-03-08-23-01"
      },
      "protocolDigest": "sha256:a79eaf4df66c97c6adc1fb2d09df78682155d1ea86652ae0b70199875bde7747",
      "workloadDigest": "sha256:493132b52fe67713ac3857847cf8c64bbbbd0a7ef3396a8062a95e5790ef6396",
      "environmentDigest": "sha256:2750ae381d6de582f1eb6b9008e0987463d19ede8af127d97d76c5ae09121393",
      "implementationDigest": "sha256:df6427ef8ed7d7d4f56c47ca0252c3ea15dc4283d04093de0ff6853aff06c2d0"
    },
    "metadata": {
      "name": "liger-kto-loss-forward-huggingface-nvidia-h100-80gb-hbm3-batch4-hidden1024-seq512-vocab128256-bf16-2",
      "title": "Fused linear + KTO loss · huggingface · forward",
      "labels": {
        "liger_version": "0.5.4"
      }
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  },
  "protocol": {
    "kind": "BenchmarkProtocol",
    "spec": {
      "harness": {
        "name": "Liger-Kernel benchmark scripts",
        "repository": "https://github.com/linkedin/Liger-Kernel"
      },
      "measurement": {
        "timer": "cuda_events",
        "primaryStatistic": "median"
      },
      "comparability": {
        "notes": "Timed: the module forward pass, via triton.testing.do_bench with quantiles 0.5/0.2/0.8 (median with p20/p80 spread). The upstream CSV records no CUDA, driver, or torch version; the Liger release rides each run's labels. Comparable only within one kernel, workload, timed pass, and GPU.",
        "family": "liger_kernel_bench"
      }
    },
    "metadata": {
      "name": "liger-bench-do-bench-forward",
      "title": "Liger benchmark harness · forward pass"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  },
  "environment": {
    "kind": "ExecutionEnvironment",
    "spec": {
      "hardware": {
        "vendor": "nvidia",
        "product": "NVIDIA H100",
        "formFactor": "80GB HBM3",
        "architecture": "sm_90"
      },
      "software": {}
    },
    "metadata": {
      "name": "liger-bench-nvidia-h100-80gb-hbm3",
      "title": "Liger benchmark host · H100 80GB HBM3"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  }
}

Cite this record (permalink, digest, access date)

Report an issue with this run

Published 2026-08-20 · Liger-Kernel benchmarks · BSD-2-ClauseAll results for Fused linear + KTO loss →JSON