PyTorch

{
  "run": {
    "kind": "BenchmarkRun",
    "spec": {
      "status": "passed",
      "timing": {
        "latencyNs": {
          "median": 2290080
        },
        "primaryStatistic": "median"
      },
      "observedAt": "2026-05-27T17:14:29.000Z",
      "measurements": [
        {
          "unit": "ns",
          "value": 2288390,
          "metric": "latency",
          "statistic": "p20"
        },
        {
          "unit": "ns",
          "value": 2290630,
          "metric": "latency",
          "statistic": "p80"
        }
      ],
      "sourceNative": {
        "source": "liger-kernel-bench",
        "metrics": {
          "ms_20": 2.2883904933929444,
          "ms_50": 2.2900800704956055,
          "ms_80": 2.2906303882598875
        },
        "benchmark": "jsd/forward",
        "externalId": "jsd/forward/torch/nvidia-b200/tokens1024-vocab128256-fp32/2026-05-27-17-14-29"
      },
      "protocolDigest": "sha256:a79eaf4df66c97c6adc1fb2d09df78682155d1ea86652ae0b70199875bde7747",
      "workloadDigest": "sha256:7c779cf7e97b538c512b7f64fe9674d473cae01e534b7f97814bcc29480578bc",
      "environmentDigest": "sha256:ad459e643ef802f7008f1fa8915785c905d05abbfdb5eb0328be214f8d9f3a8b",
      "implementationDigest": "sha256:a23886e9eb82a5ee86dc8da4a8b71c3d32e2af6c350b02bb1702dd8d8bdb05b1"
    },
    "metadata": {
      "name": "liger-jsd-forward-torch-nvidia-b200-tokens1024-vocab128256-fp32-2026-05-27-17-14-29",
      "title": "Jensen-Shannon divergence loss · torch · forward",
      "labels": {
        "liger_version": "0.8.0"
      }
    },
    "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 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 B200",
        "architecture": "sm_100"
      },
      "software": {}
    },
    "metadata": {
      "name": "liger-bench-nvidia-b200",
      "title": "Liger benchmark host · B200"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  }
}