onnx-llama2_xs_460M_experimental / quantize_config.json
Felladrin's picture
Add models files
737676c
{
"per_channel": false,
"reduce_range": false,
"per_model_config": {
"decoder_model": {
"op_types": [
"Where",
"Sigmoid",
"ConstantOfShape",
"Unsqueeze",
"Sub",
"Expand",
"Squeeze",
"Add",
"ReduceMean",
"Constant",
"MatMul",
"Range",
"Neg",
"Sqrt",
"Transpose",
"Shape",
"Less",
"Concat",
"Mul",
"Pow",
"Reshape",
"Cast",
"Div",
"Slice",
"Equal",
"Softmax",
"Gather"
],
"weight_type": "QInt8"
},
"model": {
"op_types": [
"Where",
"Sigmoid",
"ConstantOfShape",
"Unsqueeze",
"Sub",
"Expand",
"Squeeze",
"Add",
"ReduceMean",
"Constant",
"MatMul",
"Range",
"Neg",
"Sqrt",
"Transpose",
"Shape",
"Less",
"Concat",
"Mul",
"Pow",
"Reshape",
"Cast",
"Div",
"Slice",
"Equal",
"Softmax",
"Gather"
],
"weight_type": "QInt8"
}
}
}