AI Tools
How AirLLM runs 70 billion parameter models on GPUs with a few gigabytes of memory
The open-source AirLLM library loads a model one layer at a time so that very large open models can run on consumer graphics cards, at the cost of speed.
The open-source AirLLM library loads a model one layer at a time so that very large open models can run on consumer graphics cards, at the cost of speed.