Why does quantized model parameters return an empty list?

quantization
1

I’m trying to inspect the parameters of a quantized model to see how they differ from a regular, non-quantized model. However, the .parameters() call on a quantized model returns an empty list:

qmodel = torch.quantization.quantize_dynamic(MyCell(), {torch.nn.Linear}, dtype=torch.qint8)

list(qmodel.parameters())

returns:

[]

Is this expected? Is this because they don’t expect the weights to be trained anymore?

How can I see what the weight / bias values are?

2

yeah the weights are not trainable anymore, you can print the qmodel and look at the structure of the model and print the weight/bias for linear layers I think