roberta-base-go_emotions-onnx / quantize_config.json
smart4lec's picture
onnx release
d5af851
{
"per_channel": true,
"reduce_range": true,
"per_model_config": {
"model": {
"op_types": [
"Add",
"Erf",
"MatMul",
"Constant",
"ConstantOfShape",
"CumSum",
"Softmax",
"Slice",
"Shape",
"Concat",
"Reshape",
"Expand",
"Equal",
"Gather",
"Transpose",
"Sub",
"Cast",
"Gemm",
"Tanh",
"Where",
"Sqrt",
"Mul",
"Not",
"Div",
"ReduceMean",
"Pow",
"Unsqueeze"
],
"weight_type": "QInt8"
}
}
}