Tool/software:
Hello, I am using the sample code given by our company for quantitative perception training (QAT), and I have a problem in the code:
I have seen the quantitative tools provided by your company: It is mentioned in Github Readme:
Optional: We have provided a utility function called torchvision.edgeailite.xnn.utils.load_weights() that prints which parameters are loaded correctly and which are not - you can use this load function if needed to ensure that your parameters are loaded correctly.
. I want to use torchvision.edgealite.xnn.utils.load_weights() tooling code for quantitative perception model checking after the training, but, I see the sample code given the following code:
# load pretrained model
if pretrained_data is not None and not is_onnx_model:
model_orig = get_model_orig(model)
for (p_data,p_file) in zip(pretrained_data, pretrained_files):
print("=> using pretrained weights from: {}".format(p_file))
if hasattr(model_orig, 'load_weights'):
model_orig.load_weights(pretrained=p_data, change_names_dict=change_names_dict)
else:
xnn.utils.load_weights(get_model_orig(model), pretrained=p_data, change_names_dict=change_names_dict)
With the above code, I can't directly execute the xnn.utils.load_weights() function in else.
Do I need to modify other codes or parameters? Or delete if, else, and execute the xnn.utils.load_weights() function directly?
Looking forward to your reply!
