LLM training in simple, pure C/CUDA (opens in new tab)

(twitter.com)

4 pointstheaniketmaurya2y ago1 comments

1 comments

1 comments · 1 top-level

LLM training in simple, pure C/CUDA. There is no need for 245MB of PyTorch or 107MB of cPython.

j / k navigate · click thread line to collapse