Andre Slavescu

*Description of changes:* - added rich (version==13.3.2) to requirements.txt (experienced dependency conflict without) - added z-loss

GPT-J model

6

reference to issue https://github.com/vllm-project/vllm/issues/198

[Suggestion] Discussions tab for general help

2

Just as a simple suggestion, I'm wondering if there can be a discussion tab to post problems with running the system as opposed to posting individual issues. This can keep...

gelu implementation. process in chunks + usage of hardware intrinsics for slightly faster performance. Results: (original implementation) GPU cuda_gelu execution times: [13.78, 13.26, 13.95, 13.0, 12.93, 13.46] (second implementation) GPU...

Adding More Optimizers

2

## Goal: - decide which optimizers are worth implementing in a priority order ## Remaining Optimizers + General Usage from my understanding there are really only 2 from the ones...

ThunderKittens Backend

1

Would it be an idea to define the same kernels that exist in the CUDA backend with ThunderKittens as well? They have cool examples with FlashAttention2 and I think it...

Optimize Platform + Restructure

3

- [x] Restructure Makefile (automate detection of compute capability) - [ ] Optimize existing kernels

ref: https://github.com/castorini/ura-projects/issues/39