Jacob Prince-Bieker

There might be discontinuities and by default there isnt't anything to merge the patches. Although with a large enough cluster, you probably could just predict for the whole of a...

Yeah, based off the discussion you had, I think changing the default forward pass to take in which lead time makes the most sense, would reduce memory issues quite a...

I've started making my training script, and changed MetNet and MetNet2 to only do one lead time per forward pass in #20

> I agree that this is not an issue. But how do the users set 8 axial attention layers (4 x-direction and 4 y-direction.色separate one another)? That would be with...

The PR in MetNet showing the errors is here #2

> Thanks a lot for your prompt reply and comment. Would try to use those available in the huggingface first. See if I can make the model run. Hi, just...

Related to #65

Oh interesting! Yeah, I'll update this issue when I try it out to see if it makes a difference. Still have been working out a few kinks in the data...

Actually, this is EUMETSAT data! Including some of the optimum cloud masks, and I am assuming the 15 minute full disk images. So for transfer learning this would actually probably...

This library also seems like a cool way to easily try out BlurPool, and other things with Torch models: https://github.com/mosaicml/composer