EleutherAI/nanoGPT-mup
Implements maximal update parameterization technique
View on index · View in 3D Map
Related repos
- competes to train GPT-2 models super fast on 8xH100 GPUs using speedrun benchmarks
- Code for training and fine-tuning GPT language models
- Parameter-efficient fine-tuning algorithms for Stable Diffusion
- OpenAI-compatible proxy that boosts LLM reasoning accuracy with 20+ optimization techniques, no retraining needed
- benchmarking CLI commands with hyperfine
- Mines and analyzes version-control data
- Large-scale MoE model activating 14B parameters
- Produces stable timestamps for Whisper
- Optimizes macOS via the command line
- Fast and small JavaScript tokenizer for OpenAI models
