KellerJordan/modded-nanogpt
competes to train GPT-2 models super fast on 8xH100 GPUs using speedrun benchmarks
View on index · View in 3D Map
Related repos
- Code for training and fine-tuning GPT language models
- GPU kernels written from scratch
- Implements maximal update parameterization technique
- Model quantization with hardware acceleration for Nvidia, AMD, and Intel
- open source video generation for low VRAM GPUs
- Python framework for fast GPU simulation and graphics
- End-to-end multimodal LLM for any-to-any I/O
- Lightweight 2D graphics library for modern GPUs
- Fast and small JavaScript tokenizer for OpenAI models
- fine-tuning diffusion transformers for long-range video style transfer and context
