gpt2-medium-onnx-js-quantized / quantize_config.json
dmmagdal's picture
Upload 14 files
f4000fb
{
"per_channel": false,
"reduce_range": false,
"per_model_config": {
"decoder_with_past_model": {
"op_types": [
"Shape",
"Sqrt",
"Constant",
"Where",
"Mul",
"Slice",
"ReduceMean",
"Gemm",
"ConstantOfShape",
"Squeeze",
"Div",
"Unsqueeze",
"MatMul",
"Range",
"Softmax",
"Split",
"Transpose",
"Gather",
"Cast",
"Pow",
"Sub",
"Tanh",
"Add",
"Reshape",
"Concat"
],
"weight_type": "QInt8"
},
"decoder_model": {
"op_types": [
"Shape",
"Sqrt",
"Constant",
"Where",
"Mul",
"Slice",
"ReduceMean",
"Gemm",
"ConstantOfShape",
"Squeeze",
"Div",
"Unsqueeze",
"MatMul",
"Range",
"Softmax",
"Split",
"Transpose",
"Gather",
"Cast",
"Pow",
"Sub",
"Tanh",
"Add",
"Reshape",
"Concat"
],
"weight_type": "QInt8"
},
"decoder_model_merged": {
"op_types": [
"Shape",
"Sqrt",
"Constant",
"Where",
"Mul",
"Slice",
"ReduceMean",
"Gemm",
"ConstantOfShape",
"Squeeze",
"Div",
"Unsqueeze",
"MatMul",
"Range",
"Softmax",
"If",
"Split",
"Transpose",
"Gather",
"Cast",
"Pow",
"Sub",
"Tanh",
"Add",
"Reshape",
"Concat"
],
"weight_type": "QInt8"
}
}
}