PRPA1984

Results 2 issues of PRPA1984

When loading Flux model, the entire model is being created before the cast. ```python model = Flux(configs[name].params).to(torch.bfloat16) ``` The issue here is that a lot of RAM is being drained...

According to the README.md, the required Python version is >= 3.7. I think this is wrong, because the used type annotations are available in Python >= 3.9. Here is an...