How to apply torch.quantization.quantize_dynamic for conv2d layer?

[Midhilesh29]

I am working on quantizing resnet50 model. I tried to use the following command.

quantized_model = torch.quantization.quantize_dynamic(
    resnet18, {torch.nn.Conv2d,torch.nn.Linear}, dtype=torch.qint8
)

But only the linear layer has quaantized but not the convolutional layer. Can anyone help me how to dynamically quantize the convolutional layer?

[Willert98]

same problem,have some solution?

[kamata1729]

It seems dynamic quantization does not supports conv2d currently
https://discuss.pytorch.org/t/cannot-quantize-nn-conv2d-with-dynamic-quantization/66722/2

[CoinCheung]

Hi,

in the tutorials, the batchnorm layers are always fused into conv layers. Does this mean that quantization of batchnorm layers are not supported also, even with static quantization ?