OBrienLabs Dev
OBrienLabs Corporation
Pinned Loading
Repositories
Showing 10 of 53 repositories
-
- performance Public
Systematically determine the optimized spot where we can push our hardware to the fullest possible use. Multithreaded optimization depends on multiple factors including CPU/GPU type (M4Max vs 14900 or MetalCUDA. Operations involving space-time tradeoffs like heap usage need to be fine tuned around batch sizes.
- foundation-transformer-llm Public
Training a base transformer model to be used as a foundation model from first principles
Top languages
Loading…
Most used topics
Loading…