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
4.27k stars 228 forks source link

fix sentence-filter adding separator bug and add document #102

Closed SiyunZhao closed 4 months ago

SiyunZhao commented 4 months ago

fix sentence-filter adding separator bug and add readme for structured_compress_prompt