AlphaUncommon
tysam-code/hlb-gpt
Minimalistic, extremely fast, and hackable researcher's toolbench for GPT models in 307 lines of code. Reaches <3.8 validation loss on wikitext-103 on a single A100 in <100 seconds. Scales to larger models with one parameter change (feature currently in alpha).
- Owner
- tysam-code (personal account)
- Stars
- Stars: 360
- Forks
- Forks: 27
- Language
- Python
- License
- Apache-2.0
- Created
- 2023
- Tier
- Alpha (250 to 699 ★) What this tier means
More repos by tysam-code
All 2 repos by tysam-code- tysam-code/hlb-CIFAR10Stars: 1,317PythonBeta
Nearby in Python
All Python repos- Neeeophytee/finding-unknowns-skillsStars: 360PythonAlpha
- chiba233/atm10-zh-cnStars: 360PythonAlpha
- Cygnux-Labs/TracekitStars: 360PythonAlpha
- jcarbaugh/django-wordpressStars: 360PythonAlpha
- viisual/ASCII-DecoratorStars: 360PythonAlpha
- f-droid/fdroiddataStars: 360PythonAlpha