Tim Dettmers @Tim_Dettmers We release LLM.int8(), the first 8-bit inference method that saves 2x memory and does not degrade performance for 175B models by exploiting emergent properties. Read More: Paper: https://arxiv.org/abs/2208.07339 Software: https://huggingface.co/blog/hf-bitsandbytes-integration… Emergence: https://timdettmers.com/2022/08/17/llm-int8-and-emergent-features/