I went through the documentation and blogs on pytorch but did not find a way to compare the model before and after quantisation (specially the weight compression aspect of it). It would also be really helpful if the method to compare works on custom quantisation implementations as this is something I desire.
Could you, please, provide an example of what you are doing?
There are some ways to see the effects: You can use the NS to see what happens to the weights during the quantization or you could just save the model and compare the sizes.
Hey. I am trying to produce a quantized VIT model for image classification. I wanted tools for comparing the difference in space complexities and time required for the forward propagation of the model. I have figured it out though , so thank you! I will also look into NS, seems like something that would be helpful for similar tasks. Good day!