{
  "domain": "com.microsoft",
  "name": "Gelu",
  "sinceVersion": 1,
  "description": "Applies the Gaussian Error Linear Unit (GELU) activation elementwise: `Y = 0.5 * X * (1 + erf(X / sqrt(2)))`. The output has the same shape as the input. Float16 and float32 are supported; the schema's double and bfloat16 types are not.",
  "inputs": [
    { "role": "X", "dtype": "T", "description": "Values transformed elementwise by the exact GELU activation." }
  ],
  "outputs": [
    {
      "role": "Y",
      "dtype": "T",
      "rank": "ranks.X",
      "shape": "shapes.X",
      "description": "Output tensor after applying GELU; same shape as the input."
    }
  ],
  "typeConstraints": { "T": ["float32", "float16"] },
  "args": {
    "X": { "kind": "tensor", "semantic": "X", "role": "input" },
    "Y": { "kind": "tensor", "semantic": "Y", "role": "output" }
  },
  "tunables": { "WORKGROUP_SIZE": 256 },
  "constants": {
    "scalar": "dtypes.T",
    "usesF16": "dtypes.T == \"f16\"",
    "approximate": "\"erf\"",
    "vec4Tail": false,
    "hasBias": false
  },
  "variants": [
    {
      "id": "vec4",
      "priority": 20,
      "when": ["numel(shapes.X) == numel(shapes.Y)", "f16Ok(dtypes.T)", "numel(shapes.X) > 0", "numel(shapes.X) % 4 == 0"],
      "constants": { "vec4": true, "vectorScalar": "\"vec4<\" ~ dtypes.T ~ \">\"" },
      "passes": [
        {
          "id": "main",
          "name": "Gelu.vec4",
          "shader": "elementwise-bias-gelu.wgsl.jinja",
          "bindings": [
            {
              "name": "x",
              "arg": "X",
              "semantic": "X",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$vectorScalar"
            },
            {
              "name": "y",
              "arg": "Y",
              "semantic": "Y",
              "buffer": { "type": "storage" },
              "elementType": "$vectorScalar"
            },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": {
                "name": "Params",
                "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.X) / 4" }]
              }
            }
          ],
          "dispatch": { "threads": "numel(shapes.X) / 4", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ]
    },
    {
      "id": "scalar",
      "priority": 0,
      "when": ["numel(shapes.X) == numel(shapes.Y)", "f16Ok(dtypes.T)"],
      "constants": { "vec4": false },
      "passes": [
        {
          "id": "main",
          "name": "Gelu.scalar",
          "shader": "elementwise-bias-gelu.wgsl.jinja",
          "bindings": [
            {
              "name": "x",
              "arg": "X",
              "semantic": "X",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            { "name": "y", "arg": "Y", "semantic": "Y", "buffer": { "type": "storage" }, "elementType": "$scalar" },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": { "name": "Params", "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.X)" }] }
            }
          ],
          "dispatch": { "threads": "numel(shapes.X)", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ]
    }
  ]
}