How the "vgg_conv.pth" is generated?

[panovr]

I found that "vgg_conv.pth" only has the conv layers of the original vgg net, and the fc layers are removed. May I ask that how this model is generated?

[leongatys]

• Load the full model e.g. from the caffe model zoo
• Define the network architecture in pytorch that only includes the convolutional part.
• Copy the numpy arrays of the parameters for the convolutional layers from the full caffe model into the pytorch model.
• Save the state dict of the pytorch model.

Hope that helps, Leon

[panovr]

Got it, thanks!

[michaelhuang74]

@leongatys Thanks for the steps. Could you kindly share the code that converts the caffe model to pytorch model?