PyTorch

{
  "run": {
    "kind": "BenchmarkRun",
    "spec": {
      "status": "passed",
      "timing": {
        "latencyNs": {
          "median": 2554080
        },
        "primaryStatistic": "median"
      },
      "observedAt": "2026-07-26T00:09:58.000Z",
      "measurements": [
        {
          "unit": "ns",
          "value": 2553056,
          "metric": "latency",
          "statistic": "p20"
        },
        {
          "unit": "ns",
          "value": 2555040,
          "metric": "latency",
          "statistic": "p80"
        }
      ],
      "sourceNative": {
        "source": "liger-kernel-bench",
        "metrics": {
          "ms_20": 2.553056001663208,
          "ms_50": 2.554080009460449,
          "ms_80": 2.555039882659912
        },
        "benchmark": "megatron_swiglu/full",
        "externalId": "megatron_swiglu/full/megatron/nvidia-h100-80gb-hbm3/batch4-ffn16384-seq2048-bf16/2026-07-26-00-09-58"
      },
      "protocolDigest": "sha256:5b02ee9a2140fbc6fcc1c89800f6c177ba22718c8897cb532bf0ba91867c1e58",
      "workloadDigest": "sha256:acee3739a61c6b89ba65dd0e3d4d2aa8866008c022b00aff3a14c1d5f6a46e5c",
      "environmentDigest": "sha256:2750ae381d6de582f1eb6b9008e0987463d19ede8af127d97d76c5ae09121393",
      "implementationDigest": "sha256:ebdec3855511a3b6ae7df50fa27a70da8c06b0eb5d8026a2fe27779438c068bb"
    },
    "metadata": {
      "name": "liger-megatron-swiglu-full-megatron-nvidia-h100-80gb-hbm3-batch4-ffn16384-seq2048-bf16-2026-07-26-00",
      "title": "SwiGLU activation, Megatron layout · megatron · full",
      "labels": {
        "liger_version": "0.8.1"
      }
    },
    "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 forward pass plus the backward 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-full",
      "title": "Liger benchmark harness · full 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"
  }
}