bart-large-mnli / quantize_config.json
matevladimir's picture
Upload folder using huggingface_hub
d0e94a1 verified
raw
history blame
993 Bytes
{
"per_channel": true,
"reduce_range": true,
"per_model_config": {
"model": {
"op_types": [
"Div",
"Concat",
"Slice",
"Where",
"Sqrt",
"ScatterND",
"ReduceMean",
"Constant",
"Less",
"Equal",
"Add",
"Cast",
"Tanh",
"Unsqueeze",
"Softmax",
"Erf",
"MatMul",
"Range",
"Squeeze",
"Shape",
"Gather",
"Mul",
"NonZero",
"Transpose",
"Expand",
"Pow",
"ConstantOfShape",
"GatherND",
"Gemm",
"Sub",
"Reshape"
],
"weight_type": "QInt8"
}
}
}