lyogavin / airllm

AirLLM 2.8T LLM inference with single 4GB GPU

higgsfield-ai / higgsfield

Fault-tolerant, highly scalable GPU orchestration, and a machine learning framework designed for training models with billions to trillions of parameters