Sebastian Raschka
Sebastian Raschka
> But if we don't want anybody having to write any code then we should have it. For my part, I actually like seeing the text generation for a fixed...
@a-r-j This is super cool. Btw. perhaps we don't need to worry about it extra dependencies here because NumPy already uses Cython (https://github.com/numpy/numpy/blob/main/build_requirements.txt), and pandas is build on NumPy, and...
Ah sorry for the incomplete code. Yeah, it seems weird, I am not getting any warnings, either. I think we can close this and sorry for bothering with that!
Arg, yes, this should be `--data.val_split_fraction 0.1`. Regarding the permission error: That's a weird one. Could you try to create the folders manually before trying `litgpt download --repo_id microsoft/phi-2`. I...
> Can repo support GaLore soon ? In the works via #1192 !
Thanks for sharing and writing-up this thorough description. I saw the paper a few months back but must admit that I didn't have time to read. Btw I am all...
To me both would be welcome and valuable contributions :). I would maybe do both in separate PRs as that would make the code review a bit easier
Just ran a full Alpaca finetuning round for StableLM 3B and the results looks great. Loss converges to ~1 and it generates very sensible outputs. It takes 45 min on...
Thanks for the review, I will try to address these cases tonight / tomorrow morning. Btw a question regarding the CI. It looks like it got automatically canceled, or is...
> Uh that's strange. Try pushing new commits and I'll debug it if it keeps happening Seems to work now, no worries.