chadnorvell
Follow
Stars
2
stars
written in Python
Clear filter
To speed up LLMs' inference and enhance LLM's perceive of key information, compress the prompt and KV-Cache, which achieves up to 20x compression with minimal performance loss.