Haydn Jones

I find it easy enough to install as well, but I would appreciate a Snap package.

I also have my runs crash a surprising amount due to the dataloader crashing because of the hub, some way to address this would be nice.

> I think this kind of optimisation should happen automatically at the XLA level. Outside of the jit'd region, store your input in NHWC. Then pass into the jit'd region,...

I was testing one with ~14 million parameters and was running train batches at ~20it/s.

Yes we can *certainly* support that, but likely not on the web version. ONNXRuntime Web's support for GPU's isn't great. I can write a python script using PyTorch that will...

> yes, thats perfect! Great. I'll jot down some features I'll consider adding: - [ ] Command line configuration - [ ] Single image input support - [ ] Support...

Just bumping this to let you know I'm still working on this. Really just need to get the CLI set up and stuff, inference is done.

We'll get this out after we finish the website rewrite/redesign (you can get a sneak peek here https://de-spaghetti.waifuxl.pages.dev/). We're pretty close!

This stupid bot keeps closing this issue. I'll get it done. Thanks for the support everyone

This will be addressed when we do 2.0. Pinky promise