EfficientQAT: LLM Quantization, gets a 2-bit llama2-70B outperform regular 13Bold.reddit.com21 points·jackbravo··0 commentsOpen articleSaveView on HN