AWQ

github.com ↗

- Activation-aware Weight Quantization for LLM Compression and Acceleration.