AWQ: Activation-Aware Weight Quantization¶
Authors: Lin et al. Year: 2023 ArXiv/Link: https://arxiv.org/abs/2306.00978
Summary¶
Improved quantization by considering activation statistics, protecting important weights from aggressive quantization.
Key Concepts¶
- Activation-aware
- Weight quantization
- Statistical awareness
- Quality preservation
- Inference optimization
Impact¶
Better quantization quality than weight-only methods
Category¶
Quantization