AirLLM is a model-serving approach and Python package that lets very large language models run with far less VRAM, including 70B-class models on a single 4GB GPU in the project’s examples and updates.
Sign in to read the manager-framed brief.
Sign inAI-explained · grounded in each repo's README