Convert .pytorch model to .ckpt?

[deama]

I was wondering if there is a .ckpt version available of the "SepConv - L_F" model, or perhaps if there is a way to convert the .pytorch model to a .ckpt one?

[avinashpaliwal]

.ckpt is just an extension used to save the model, you can use .pth as well. I think as long as the checkpoint was created using the torch.save(), you can load it back using torch.load().

[deama]

So if I change "network-lf.pytorch" to "network-lf.ckpt" it should work?

[avinashpaliwal]

There is no need to rename the file. Looking at his code here, you can load the checkpoint like this:

model.load_state_dict(torch.load('checkpointName.pytorch'))

[deama]

Sorry, where do I put that?

[avinashpaliwal]

@sniklaus @deama Are you trying to run the SepConv CNN? Then just follow the usage instructions provided on the project page.

[sniklaus]

It looks like @deama is trying to load the SepConv weights into the Super Slo-Mo model. I am afraid that this is not possible, they are two different methods that are not compatible with each other. If you want to run SepConv, then I concur with @avinashpaliwal that you will have to follow the instructions in the SepConv repository: https://github.com/sniklaus/pytorch-sepconv

[deama]

Hmm, I see.