gitmyhub

exllama

Python ★ 2.9k updated 2y ago

A more memory-efficient rewrite of the HF transformers implementation of Llama for use with quantized weights.

No plain-English explanation yet — one is being written right now. Check back in a minute.