Skip to content

AirLLM 2.8T LLM inference with single 4GB GPU

261 B
Raw Blame
{"text": "AirLLM streams frozen decoder weights from disk one layer at a time so LoRA can train on a small GPU."}
{"prompt": "What is AirLLM?", "completion": "A library that runs and trains huge language models on small VRAM by streaming one layer at a time."}