ort_config.json

{
  "one_external_file": true,
  "opset": null,
  "optimization": {},
  "quantization": {
    "activations_dtype": "QUInt8",
    "activations_symmetric": false,
    "format": "QOperator",
    "is_static": false,
    "mode": "IntegerOps",
    "nodes_to_exclude": [],
    "nodes_to_quantize": [],
    "operators_to_quantize": [
      "Concat",
      "Constant",
      "Resize",
      "Conv",
      "Div",
      "Unsqueeze",
      "Shape",
      "Gemm",
      "MatMul",
      "Transpose",
      "Pow",
      "Softmax",
      "Slice",
      "Where",
      "Expand",
      "ReduceMean",
      "Mul",
      "Add",
      "Sqrt",
      "Gather",
      "Sub",
      "ConstantOfShape",
      "Reshape",
      "Equal",
      "Identity",
      "Erf"
    ],
    "per_channel": true,
    "qdq_add_pair_to_weight": false,
    "qdq_dedicated_pair": false,
    "qdq_op_type_per_channel_support_to_axis": {
      "MatMul": 1
    },
    "reduce_range": true,
    "weights_dtype": "QUInt8",
    "weights_symmetric": true
  },
  "use_external_data_format": false
}