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.