search

microsoft / LLMLingua

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.
https://llmlingua.com/
MIT License
4.18k stars 222 forks source link

add feature: compress_json #120

Closed SiyunZhao closed 3 months ago

SiyunZhao commented 3 months ago