The minimum data type set

Abierto

#853 abierto el 28/5/2025 6 comentarios 4 reacciones 0 asignados Ver en GitHub

Nadie ha tomado este issue todavía.

Evaluación

Dificultad: 5/5
Tiempo estimado: Más de una semana
Aptitud para principiantes: 12/100

Tipo de issue: Nueva funcionalidad
Claridad: Necesita aclaración
Estado de actividad: Estancado
Área: machine-learning

Línea de trabajo

This is a spec-level decision about the minimum data type set, not a code change. Read the linked WebML minutes and the referenced webnn issues #278 and #853's comment thread, then compare the opSupportLimits intersection JSON against the spec's data type tables. The open questions are whether l2Pool2d (missing in TFLite) and sign (missing in CoreML) block the minimum set, and the work is done when the group records a decision on that set.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

operator specific

As a follow-up for the discussion at WebML WG Teleconference – 22 May 2025, I checked the data types (and rank range) of all ops via opSupportLimits across DirectML, TFLite and CoreML backends in Chromium prototype.

The result is promising: all ops have data type overlap except two ops:

l2Pool2d: it is not implemented in TFLite backend yet (Chromium issue), because of a TFLite issue which is discussed in https://github.com/webmachinelearning/webnn/issues/278#issuecomment-2887950826. Once it is fixed, the overlap should be "float32" and "float16"
sign: it is not implemented the CoreML backend yet (Chromium issue). CoreML sign supports fp16, fp32, i32. Once it is supported, the input data types intersection should be "float32", "float16" and "int32".

Please find the details in the following opSupportLimits intersection JSON dump. /cc @reillyeon @fdwr

The opSupportLimits intersection of DirectML, TFLite and CoreML backends

The JSON output

{
  "abs": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "add": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "argMax": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "int32" ]
    }
  },
  "argMin": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "int32" ]
    }
  },
  "averagePool2d": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "batchNormalization": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 5 }
    },
    "mean": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    },
    "scale": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "variance": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    }
  },
  "cast": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "ceil": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "clamp": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "concat": {
    "inputs": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "constant": {
    "dataTypes": [ "float32", "float16", "int32", "uint32", "int8", "uint8", "int4" ]
  },
  "conv2d": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "filter": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "convTranspose2d": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "filter": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "cos": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "cumulativeSum": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "dequantizeLinear": {
    "input": {
      "dataTypes": [ "int32", "int8", "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32" ]
    },
    "scale": {
      "dataTypes": [ "float32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "zeroPoint": {
      "dataTypes": [ "int32", "int8", "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    }
  },
  "div": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "elu": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "equal": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "erf": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "exp": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "expand": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "floor": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "gather": {
    "indices": {
      "dataTypes": [ "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "gatherElements": {
    "indices": {
      "dataTypes": [ "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "gatherND": {
    "indices": {
      "dataTypes": [ "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "gelu": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "gemm": {
    "a": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "b": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "c": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 2 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "greater": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "greaterOrEqual": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "gru": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "initialHiddenState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "outputs": {
      "dataTypes": [ "float32", "float16" ]
    },
    "recurrentBias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "recurrentWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "weight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    }
  },
  "gruCell": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "hiddenState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    },
    "recurrentBias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "recurrentWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "weight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    }
  },
  "hardSigmoid": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "hardSwish": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "identity": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "input": {
    "dataTypes": [ "float32", "float16", "int32" ]
  },
  "instanceNormalization": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    },
    "scale": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    }
  },
  "l2Pool2d": {
    "input": {
      "dataTypes": [],
      "rankRange": {}
    },
    "output": {
      "dataTypes": []
    }
  },
  "layerNormalization": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    },
    "scale": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    }
  },
  "leakyRelu": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "lesser": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "lesserOrEqual": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "linear": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "log": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "logicalAnd": {
    "a": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "logicalNot": {
    "a": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "logicalOr": {
    "a": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "logicalXor": {
    "a": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "lstm": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "initialCellState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "initialHiddenState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "outputs": {
      "dataTypes": [ "float32", "float16" ]
    },
    "peepholeWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "recurrentBias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "recurrentWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    },
    "weight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 3, "max": 3 }
    }
  },
  "lstmCell": {
    "bias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "cellState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "hiddenState": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "outputs": {
      "dataTypes": [ "float32", "float16" ]
    },
    "peepholeWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "recurrentBias": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 1 }
    },
    "recurrentWeight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    },
    "weight": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 2 }
    }
  },
  "matmul": {
    "a": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 2, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "max": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "maxPool2d": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "min": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "mul": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "neg": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "notEqual": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "uint8" ]
    }
  },
  "output": {
    "dataTypes": [ "float32", "float16", "int32" ]
  },
  "pad": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "pow": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "prelu": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    },
    "slope": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 4 }
    }
  },
  "quantizeLinear": {
    "input": {
      "dataTypes": [ "float32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "int8", "uint8" ]
    },
    "scale": {
      "dataTypes": [ "float32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "zeroPoint": {
      "dataTypes": [ "int8", "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    }
  },
  "reciprocal": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceL1": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "reduceL2": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceLogSum": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceLogSumExp": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceMax": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "reduceMean": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceMin": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "reduceProduct": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reduceSum": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "reduceSumSquare": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "relu": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "resample2d": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 4, "max": 4 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "reshape": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "reverse": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "scatterElements": {
    "indices": {
      "dataTypes": [ "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    },
    "updates": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    }
  },
  "scatterND": {
    "indices": {
      "dataTypes": [ "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    },
    "updates": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 1, "max": 5 }
    }
  },
  "sigmoid": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "sign": {
    "input": {
      "dataTypes": [],
      "rankRange": { "min": 0, "max": 0 }
    },
    "output": {
      "dataTypes": []
    }
  },
  "sin": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "slice": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32", "int8", "uint8" ]
    }
  },
  "softmax": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 1, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "softplus": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "softsign": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "split": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "outputs": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "sqrt": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "sub": {
    "a": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "b": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "tan": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "tanh": {
    "input": {
      "dataTypes": [ "float32", "float16" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16" ]
    }
  },
  "tile": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "transpose": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "triangular": {
    "input": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 2, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    }
  },
  "where": {
    "condition": {
      "dataTypes": [ "uint8" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "falseValue": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    },
    "output": {
      "dataTypes": [ "float32", "float16", "int32" ]
    },
    "trueValue": {
      "dataTypes": [ "float32", "float16", "int32" ],
      "rankRange": { "min": 0, "max": 5 }
    }
  }
}

Lenguaje dominante: Bikeshed
Estrellas: 566
Forks: 72
Merge medio: 6 d 5 h
PR fusionados (30 d): 1

Preparar el entorno

Sin Dockerfile ni archivo de Docker Compose
Sin plantilla de pull request
Leer la guía de contribución

Primeros pasos

Lee el issue completo y luego la guía de contribución del proyecto.
Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
Haz un fork del repositorio y trabaja en una rama.
Abre un pull request que haga referencia al número del issue.

Más de webmachinelearning/webnn

Do subnormal results have to be preserved when the reference formula overflows in float32?
Abierto
- ¿Algo no está bien?
Dificultad 5/5 Más de una semana Aptitud para principiantes 35/100

webmachinelearning/webnn#967 · Abierto el 21/10/2026
Proposal: Support stateful model for WebNN
Abierto
- ¿Algo no está bien?
Dificultad 5/5 Más de una semana Aptitud para principiantes 25/100

webmachinelearning/webnn#966 · 1 comentario · Abierto el 20/10/2026
Proposal: Support build-time metadata to describe how a graph will be driven
Abierto
- ¿Algo no está bien?
Dificultad 5/5 Más de una semana Aptitud para principiantes 35/100

webmachinelearning/webnn#965 · 1 comentario · Abierto el 20/10/2026
What is the current testing/shipping status of WebNN in Chrome and Edge, and what should developers use today?
Abierto
- ¿Algo no está bien?
Dificultad 3/5 1-2 días Aptitud para principiantes 68/100

webmachinelearning/webnn#964 · 1 comentario · 1 reacción · Abierto el 20/10/2026
Proposal: Expose engine-defined compute timing for WebNN (`MLComputeTiming`)
Abierto
- ¿Algo no está bien?
Dificultad 5/5 Más de una semana Aptitud para principiantes 30/100

webmachinelearning/webnn#963 · 3 comentarios · 2 reacciones · Abierto el 20/10/2026

Todos los issues de webmachinelearning/webnn

Issues similares

[Platform][OpenAi] Non-streaming structured output parses commentary instead of final_answer
Abierto
- ¿Algo no está bien?
Bug Platform Status: Needs Review

Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100

symfony/ai#2771 · Abierto el 24/10/2026

Los mantenedores suelen responder en 1 día
reflect: analysts see no trajectories for task ids that #309 maps, so LiveMathematicianBench training makes no edits
Abierto
- ¿Algo no está bien?
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100

microsoft/SkillOpt#332 · Abierto el 24/10/2026

Los mantenedores suelen responder en 1 día
RDNA2 Navi22 (RX 6700 XT, gfx1031 run as gfx1030) on r120: prefill and decode against upstream b11804
Abierto
- ¿Algo no está bien?
discussion tuning

Dificultad 1/5 Menos de una hora Aptitud para principiantes 90/100

stew675/llama-cpp-rdna-boosts#214 · 1 comentario · Abierto el 24/10/2026

Los mantenedores suelen responder en 1 día
[Bug]: SM-43 reuses the loop variable artifact_buckets, so the EC2 instance check judges only the last role's buckets and can return Passed
Abierto
- ¿Algo no está bien?
Dificultad 2/5 1-3 horas Aptitud para principiantes 76/100

aws-samples/sample-aiml-security-assessment#90 · Abierto el 24/10/2026

Los mantenedores suelen responder en 8 días
Eval bug: minimax-m3: build_moe_ffn() drops the per-expert NVFP4 scale → garbled output (M3.1 Flash)
Abierto
- ¿Algo no está bien?
bug-unconfirmed

Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100

ggml-org/llama.cpp#31063 · Abierto el 23/10/2026

Los mantenedores suelen responder en 1 día

Más issues de Machine Learning

The minimum data type set

¿Algo no está bien?

Evaluación

Línea de trabajo

Descripción

Preparar el entorno

Primeros pasos

Más de webmachinelearning/webnn

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

Issues similares

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

¿Algo no está bien?

Recibe los nuevos issues en tu correo