HackerTrans
TopNewTrendsCommentsPastAskShowJobs

kayvr

no profile record

Submissions

Show HN: TokenHawk, WebGPU Running LLaMA

github.com
2 points·by kayvr·hace 3 años·0 comments

comments

kayvr
·el año pasado·discuss
FWIW, I thought the bits on governance and history provided much needed context. Great post.
kayvr
·hace 2 años·discuss
I've heard the WebGPU workgroup wants to close the gap on tensor core support.
kayvr
·hace 2 años·discuss
It's not too surprising that Crew Dragon is a bit like a Tesla. But, wow, the ISS has leaks on board.
kayvr
·hace 2 años·discuss
Rusty would dig this. It's the name of good ol' game.

https://en.wikipedia.org/wiki/SteamWorld
kayvr
·hace 2 años·discuss
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
·hace 3 años·discuss
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
·hace 3 años·discuss
Also https://github.com/kayvr/TokenHawk, a WebGPU implementation of LLaMA.

edit: Note that this is my project.