PhoWhisper-small / quantize_config.json
huuquyet's picture
feat: update converted models with config
515c6b5 verified
{
"per_channel": false,
"reduce_range": false,
"per_model_config": {
"encoder_model": {
"op_types": [
"Constant",
"Softmax",
"Mul",
"Sqrt",
"Erf",
"ReduceMean",
"Gather",
"Concat",
"MatMul",
"Reshape",
"Sub",
"Shape",
"Div",
"Pow",
"Transpose",
"Add",
"Unsqueeze",
"Conv"
],
"weight_type": "QUInt8"
},
"decoder_model": {
"op_types": [
"Softmax",
"Range",
"ConstantOfShape",
"Transpose",
"Gather",
"ReduceMean",
"Sub",
"Shape",
"Equal",
"Expand",
"Where",
"Sqrt",
"Erf",
"Concat",
"Reshape",
"Squeeze",
"MatMul",
"Constant",
"Slice",
"Mul",
"Cast",
"Div",
"Pow",
"Add",
"Unsqueeze",
"Less"
],
"weight_type": "QInt8"
},
"decoder_with_past_model": {
"op_types": [
"Constant",
"Slice",
"Mul",
"Gather",
"Sqrt",
"ReduceMean",
"Softmax",
"Concat",
"Erf",
"MatMul",
"Reshape",
"Sub",
"Shape",
"Div",
"Pow",
"Transpose",
"Add",
"Unsqueeze"
],
"weight_type": "QInt8"
},
"decoder_model_merged": {
"op_types": [
"Softmax",
"Range",
"ConstantOfShape",
"Transpose",
"Gather",
"ReduceMean",
"Sub",
"Shape",
"Equal",
"Expand",
"Where",
"Sqrt",
"Erf",
"Concat",
"If",
"Reshape",
"Squeeze",
"MatMul",
"Constant",
"Slice",
"Mul",
"Cast",
"Div",
"Pow",
"Add",
"Unsqueeze",
"Less"
],
"weight_type": "QInt8"
}
}
}