AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration.
Ji LinJiaming TangHaotian TangShang YangXingyu DangSong HanPublished in: CoRR (2023)
Keyphrases
- efficient compression
- quantization noise
- lossy image compression
- image compression
- uniform quantization
- compression scheme
- compression ratio
- data compression
- huffman coding
- entropy coding
- transform coding
- quantization error
- bit rate
- wavelet image coding
- bits per pixel
- adaptive quantization
- arithmetic coding
- quantization scheme
- lookup table
- compression rate
- compression algorithm
- information processing
- neural network
- weighting scheme
- lossy compression
- vector quantization
- lossless image coding
- block coding
- activation detection
- reconstructed image
- tree structured vector quantization