Benjamin Bossan

Yes, it would be nice if you could share some code. For that, I don't need the actual data, just something of the same structure, which could be used for...

Take the time that you need. On the question how: What I like to do is code up a snippet using for instance Jupyter notebook, make sure it shows the...

Hi @pgg1610 I tried to make your example work but can't train the net since it's still missing some data (`FileNotFoundError: [Errno 2] No such file or directory: '/Users/pghaneka/OneDrive -...

@zachbellay Without knowing anything about your specific case, would it be possible to have the discriminator and the generator be submodules of the same overarching module? Then the module could...

Thank you @zachbellay, I'll have a look at this as soon as I've got some time on my hands and see if it's possible to port to skorch without too...

@YannDubs Thank you for the proposal and clean explanation. > Essentially the problem comes from the fact that for GANs you need 2 models, 2 losses, 2 optimizers, 2 alternating...

> when `*Scoring` is supplied with a user-defined function (or `None` value) we _disable_ caching as a precaution and possibly notify the user about it This sounds a little bit...

I didn't know about stochastic weight averaging, thanks a lot. I looked at your code and the PyTorch example and came up with a slightly different implementation based on yours:...

@WillCastle Did you have opportunity to test this out yet? I believe it might not even be necessary store `swa_model` (`module_swa_` in my code) on the `net`, it can be...

Thanks for taking another look @WillCastle > as to multiplying by `swa_start`, I did intend it as a fraction Okay, this makes sense. I would probably allow both possibilities: if...