PyTorch

{
  "run": {
    "kind": "BenchmarkRun",
    "spec": {
      "status": "passed",
      "timing": {
        "latencyNs": {
          "median": 2855200
        },
        "primaryStatistic": "median"
      },
      "observedAt": "2024-09-04T12:59:48.000Z",
      "measurements": [
        {
          "unit": "ns",
          "value": 2852755,
          "metric": "latency",
          "statistic": "p20"
        },
        {
          "unit": "ns",
          "value": 2856454,
          "metric": "latency",
          "statistic": "p80"
        }
      ],
      "sourceNative": {
        "source": "liger-kernel-bench",
        "metrics": {
          "ms_20": 2.852755069732666,
          "ms_50": 2.8552000522613525,
          "ms_80": 2.856454372406006
        },
        "benchmark": "kl_div/full",
        "externalId": "kl_div/full/torch/nvidia-h100-pcie/tokens16384-vocab4096-fp32/2024-09-04-12-59-48"
      },
      "protocolDigest": "sha256:5b02ee9a2140fbc6fcc1c89800f6c177ba22718c8897cb532bf0ba91867c1e58",
      "workloadDigest": "sha256:656c5e61a974e5a50552c0330bd09da1bdf6af9ad9ec5a9b13bed0cf1ffeb453",
      "environmentDigest": "sha256:89f9b28b06758f22a1df331fea58bfde2c9cc0f3627d1829509743ea538f0011",
      "implementationDigest": "sha256:9d1db6030b44918b1cda264dd2cd4e22a87e96c0b0e342ebd47437ce292a270c"
    },
    "metadata": {
      "name": "liger-kl-div-full-torch-nvidia-h100-pcie-tokens16384-vocab4096-fp32-2024-09-04-12-59-48",
      "title": "KL-divergence loss · torch · full",
      "labels": {
        "liger_version": "0.2.1"
      }
    },
    "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": "PCIe",
        "architecture": "sm_90"
      },
      "software": {}
    },
    "metadata": {
      "name": "liger-bench-nvidia-h100-pcie",
      "title": "Liger benchmark host · H100 PCIe"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  }
}