Christopher Beckham, PhD

@royaljava @seungtaekoh @y-x-c sorry for a rather late addressal of the issue. I have labeled it 'enhancement request'. While we used @SinaHonari's RCN network to extract face keypoints, unfortunately it...

Hi, Sorry for the enormously long delay in response, I am sure by now my response is most likely irrelevant because you've either found the answer or moved onto something...

This was opened recently for Theano, which is exciting!: https://github.com/Theano/Theano/issues/5712 I too am interested in this since I'm working on some segmentation stuff at the moment. For the time being,...

Thanks, I will make those changes. One thing I completely forgot though... I think I would have to use an intermediate output of that layer to compute the cross-entropy, since...

Is it just me or does it seem like things could get a bit complicated for integer targets? In the conventional case, the predicted output is a tensor of shape...

Yes that makes sense. > I'm not sure how much performance differs between indexing and, err, one-hotting. I'm curious about this too.

What would be the difference between this and what's already implemented in nolearn? Is this meant to be a more 'lightweight' alternative?

Interesting, yes, I think it would be cleaner to have this as an option for `Upscale2D`. I'll have a look at that paper too.

Hi f0k, gahhh sorry for that one month delay... looks like we've both been swamped. Yes, absolutely, I'd be happy to. Let me try get back to you next week.

Having a `set_timesteps` for all relevant schedulers would also make it much easier to implement things like this: https://github.com/huggingface/diffusers/issues/7651