Static_quantization_tutorial of pytorch

Hi everyone,

I am trying to use tutorial to do model quantization but I get error when I am trying do run the line of

# Convert to quantized model
torch.quantization.convert(myModel, inplace=True)

there is error of : untimeError: Didn’t find engine for operation quantized::conv_prepack NoQEngine (operator() at /opt/conda/conda-bld/pytorch_1573049306851/work/aten/src/ATen/native/quantized/cpu/qconv_prepack.cpp:264)

BTW, I am using pytorch 3.6 env. Would be appreciated if you help me. :disappointed_relieved:

I also following the same tutorial.
Can we quantize the standard pytorch models with the same approach? How to fuse the layers if they are not given as a separate class?
Any one tried the same tools for models like VGG/ResNet ?

check what is set for - torch.backends.quantized.engine

1 Like