Run 01a0202b · KernelIndex

Primary measurement

4.17msmedian

Rank 1 in its comparison group · source-native comparison · observed 2026-04-02

ms 20 4.1689471244812015 · ms 50 4.17137598991394 · ms 80 4.173804855346679

Identity

implementationLiger

projectLiger-Kernel

revisionunknown

workloadk = 8 · hidden = 2048 · tokens = 8192 · experts = 8 · intermediate = 768 · bf16

comparison keysha256:dd8783d5bdbe6a2f…

sourceLiger-Kernel benchmarks

external idfused_moe/backward/liger/nvidia-h100-80gb-hbm3/experts8-hidden2048-intermediate768-k8-tokens8192-bf16/2026-04-02-23-59-45

sha256:55a16c48bd645cbef9154eb…

Correctness

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

Workload

hidden2048

tokens8192

experts8

intermediate768

xbf16 [8192, 2048]

down_projbf16 [8, 2048, 768]

top_k_indexint32 [8192, 8]

gate_up_projbf16 [8, 1536, 2048]

top_k_weightsbf16 [8192, 8]

definition comparatornot_asserted

Measurements

latency · median4.17 ms

latency · p204.17 ms

latency · p804.17 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": 4171376
        },
        "primaryStatistic": "median"
      },
      "observedAt": "2026-04-02T23:59:45.000Z",
      "measurements": [
        {
          "unit": "ns",
          "value": 4168947,
          "metric": "latency",
          "statistic": "p20"
        },
        {
          "unit": "ns",
          "value": 4173805,
          "metric": "latency",
          "statistic": "p80"
        }
      ],
      "sourceNative": {
        "source": "liger-kernel-bench",
        "metrics": {
          "ms_20": 4.1689471244812015,
          "ms_50": 4.17137598991394,
          "ms_80": 4.173804855346679
        },
        "benchmark": "fused_moe/backward",
        "externalId": "fused_moe/backward/liger/nvidia-h100-80gb-hbm3/experts8-hidden2048-intermediate768-k8-tokens8192-bf16/2026-04-02-23-59-45"
      },
      "protocolDigest": "sha256:080e6e66e4ec75a131aa0c9121c74b8e8eed64e7b6d621d567c6a28f79548be8",
      "workloadDigest": "sha256:2efff9b6aea730b81e8d3e09230d44bbf52960bc6bce60779315761ef968f193",
      "environmentDigest": "sha256:2750ae381d6de582f1eb6b9008e0987463d19ede8af127d97d76c5ae09121393",
      "implementationDigest": "sha256:2e70f82ea50dfe4f5eacef6799a786ced87fba66276661269d9820e9909cfce5"
    },
    "metadata": {
      "name": "liger-fused-moe-backward-liger-nvidia-h100-80gb-hbm3-experts8-hidden2048-intermediate768-k8-tokens81",
      "title": "Fused MoE expert MLP · liger · backward",
      "labels": {
        "liger_version": "0.7.0"
      }
    },
    "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 backward pass of the module output given a random upstream gradient, 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-backward",
      "title": "Liger benchmark harness · backward 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 MoE expert MLP →JSON