kayvr

kayvr

tahun laludiscuss

FWIW, I thought the bits on governance and history provided much needed context. Great post.

kayvr

2 tahun yang laludiscuss

I've heard the WebGPU workgroup wants to close the gap on tensor core support.

kayvr

2 tahun yang laludiscuss

It's not too surprising that Crew Dragon is a bit like a Tesla. But, wow, the ISS has leaks on board.

kayvr

2 tahun yang laludiscuss

Rusty would dig this. It's the name of good ol' game.

https://en.wikipedia.org/wiki/SteamWorld

kayvr

2 tahun yang laludiscuss

Integer multiplication was used to test LLMs reasoning capabilities, and I think Karpathy mentioned that tokenization might play a role in basic math. MathGLM was compared against GPT-4 in the article, but I couldn't figure out if MathGLM was trained with character-level tokenization or not.

kayvr

3 tahun yang laludiscuss

It could be. But there's quite a bit of momentum behind CUDA. Plus, CUDA is just wicked fast. I wrote a WebGPU version of LLaMA inference and there's still a bit of a gap in performance between WebGPU and CUDA. Admittedly, WebGPU can't access tensor cores and I undoubtedly need to optimize further.

kayvr

3 tahun yang laludiscuss

Also https://github.com/kayvr/TokenHawk, a WebGPU implementation of LLaMA.

edit: Note that this is my project.

Submissions

Show HN: TokenHawk, WebGPU Running LLaMA

comments