{
  "op": "ai.onnx.SimplifiedLayerNormalization",
  "cases": [
    {
      "name": "simplifiedln-f32-4096x4096",
      "preset": "smoke",
      "vars": { "dtype": "float32", "rows": 4096, "dim": 4096 },
      "attrs": { "epsilon": 0.00001, "axis": -1 },
      "inputs": {
        "x": { "shape": [4096, 4096], "dtype": "float32", "dist": "normal", "seed": 720, "scale": 0.5 },
        "scale": { "shape": [4096], "dtype": "float32", "dist": "uniform", "seed": 721, "scale": 0.25, "offset": 1 }
      },
      "outputs": { "y": { "shape": [4096, 4096], "dtype": "float32" } },
      "bench": {
        "primary": true,
        "metrics": [{ "type": "bandwidth", "value": "(args.rows * args.dim * 2 + args.dim) * dtypeBytes(args.dtype)" }]
      }
    },
    {
      "name": "simplifiedln-f32-4096x4096-stats",
      "preset": "smoke",
      "vars": { "dtype": "float32", "rows": 4096, "dim": 4096 },
      "attrs": { "epsilon": 0.00001, "axis": -1 },
      "inputs": {
        "x": { "shape": [4096, 4096], "dtype": "float32", "dist": "normal", "seed": 722, "scale": 0.5 },
        "scale": { "shape": [4096], "dtype": "float32", "dist": "uniform", "seed": 723, "scale": 0.25, "offset": 1 }
      },
      "outputs": {
        "y": { "shape": [4096, 4096], "dtype": "float32" },
        "invStdVar": { "shape": [4096, 1], "dtype": "float32" }
      },
      "bench": {
        "primary": true,
        "metrics": [
          {
            "type": "bandwidth",
            "value": "(args.rows * args.dim * 2 + args.dim) * dtypeBytes(args.dtype) + args.rows * 4"
          }
        ]
      }
    },
    {
      "name": "simplifiedln-f16-8192x3072-stats",
      "preset": "smoke",
      "vars": { "dtype": "float16", "rows": 8192, "dim": 3072 },
      "attrs": { "epsilon": 0.00001, "axis": -1 },
      "inputs": {
        "x": { "shape": [8192, 3072], "dtype": "float16", "dist": "normal", "seed": 724, "scale": 0.5 },
        "scale": { "shape": [3072], "dtype": "float16", "dist": "uniform", "seed": 725, "scale": 0.25, "offset": 1 }
      },
      "outputs": {
        "y": { "shape": [8192, 3072], "dtype": "float16" },
        "invStdVar": { "shape": [8192, 1], "dtype": "float32" }
      },
      "bench": {
        "primary": true,
        "metrics": [
          {
            "type": "bandwidth",
            "value": "(args.rows * args.dim * 2 + args.dim) * dtypeBytes(args.dtype) + args.rows * 4"
          }
        ]
      }
    }
  ]
}