Lucas Nestler

This is merely a cosmetic change to allow users to see the current download rate. Additionally this PR contains a few minor formatting changes as well as added support for...

- [ ] wait for Torch 1.11 to be released - [x] Avoid Torch 1.10 crashing - [ ] Extend ReadMe - [ ] give credit

This PR also * adds a new custom sum-based attention * changes a bunch of parameter names * changes small.yaml to integrate omnidirectional attention * breaks up our linear attention...

What we learned

6

Whenever a pull request is made, please put a summary of your knowledge in here.

Can this also run in the optimizer, or do you think it's necessary to run it within the forward pass?

[feat] patch torch and huggingface for improved performance in existing applications

1

I'd love it if we could patch existing functions like `torch.nn.functional.softmax` with our faster xformers/triton implementation. This would allow users of pre-defined models from HuggingFace to simply call `xformers.patch()` and...

What does PII mean?

3

When working with the [label prompter](https://open-assistant.io/label/label_prompter_reply), there are a couple of flags people can choose from to mark data:  While most are self-explanatory or use common phrases, I've never...