I quantize my model and I want the bit value of all weights in my model. But I don’t how to do it
Could you elaborate on what you mean by bit value of the weights?
If you would like to access quantized weights you can try <quantized_module>.weight(). This returns a quantized tensor of weights and you can use int_repr() to get the int8_t values of weights.
Based on your advice, when I try to print the weight of the quantized model, I get FP32 weights, scaling, and zero_point, but I can’t get the int weight?
yeah you can try:
to get the int weight, for more details about quantized tensor apis, please see: https://github.com/pytorch/pytorch/wiki/Introducing-Quantized-Tensor