Hello! thanks for your awesome work!
I tried your guide like #15 but the result saved file is not what I expected.
for example, if swin tiny model is 115MB, the output should be saved in around 30 MB.
and the shape of the output is torch.Size([100, 3, 224, 224]) which looks like feature map for 100 calibration data.
and the output sizes are so various per iterations
I want to save the quantized weight, not feature map.
Could you please give me some help?
Hello! thanks for your awesome work! I tried your guide like #15 but the result saved file is not what I expected. for example, if swin tiny model is 115MB, the output should be saved in around 30 MB. and the shape of the output is torch.Size([100, 3, 224, 224]) which looks like feature map for 100 calibration data. and the output sizes are so various per iterations I want to save the quantized weight, not feature map. Could you please give me some help?
Thank you so much!