PrefixQuant: Eliminating Outliers by Prefixed Tokens for Large Language Models Quantization
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916583402635264 |
|---|---|
| author | Chen, Mengzhao Liu, Yi Wang, Jiahao Bin, Yi Shao, Wenqi Luo, Ping |
| author_facet | Chen, Mengzhao Liu, Yi Wang, Jiahao Bin, Yi Shao, Wenqi Luo, Ping |
| contents | Existing weight-activation quantization methods for Large Language Models (LLMs) primarily address channel-wise outliers but often neglect token-wise outliers, which limits the accuracy of quantized models. In this work, we propose PrefixQuant, a novel quantization method that achieves state-of-the-art performance across various precision levels (W4A4KV4 and W4A8KV4) and granularities (dynamic and static quantization) by effectively isolating token-wise outliers. First, PrefixQuant eliminates token-wise outliers by prefixing outlier tokens in the KV cache, a process that is training-free and highly efficient (e.g., 1 minutes for Llama-3-70B). Second, PrefixQuant introduces new trainable parameters for block-wise training to compensate for quantization error. Our experiments show that PrefixQuant significantly outperforms existing dynamic quantization methods, even under coarser static quantization settings. For instance, PrefixQuant achieves an average accuracy improvement of +3.08 and +2.85 points over SpinQuant (dynamic quantization) on five zero-shot reasoning tasks under dynamic and static quantization settings, respectively, on W4A4KV4 Llama-3-8B. Additionally, we demonstrate up to 2.74x prefilling speedup and 2.16x decoding speedup for LLMs using W4A4 PrefixQuant. Our code is available at https://github.com/ChenMnZ/PrefixQuant. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_05265 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | PrefixQuant: Eliminating Outliers by Prefixed Tokens for Large Language Models Quantization Chen, Mengzhao Liu, Yi Wang, Jiahao Bin, Yi Shao, Wenqi Luo, Ping Machine Learning Computation and Language Existing weight-activation quantization methods for Large Language Models (LLMs) primarily address channel-wise outliers but often neglect token-wise outliers, which limits the accuracy of quantized models. In this work, we propose PrefixQuant, a novel quantization method that achieves state-of-the-art performance across various precision levels (W4A4KV4 and W4A8KV4) and granularities (dynamic and static quantization) by effectively isolating token-wise outliers. First, PrefixQuant eliminates token-wise outliers by prefixing outlier tokens in the KV cache, a process that is training-free and highly efficient (e.g., 1 minutes for Llama-3-70B). Second, PrefixQuant introduces new trainable parameters for block-wise training to compensate for quantization error. Our experiments show that PrefixQuant significantly outperforms existing dynamic quantization methods, even under coarser static quantization settings. For instance, PrefixQuant achieves an average accuracy improvement of +3.08 and +2.85 points over SpinQuant (dynamic quantization) on five zero-shot reasoning tasks under dynamic and static quantization settings, respectively, on W4A4KV4 Llama-3-8B. Additionally, we demonstrate up to 2.74x prefilling speedup and 2.16x decoding speedup for LLMs using W4A4 PrefixQuant. Our code is available at https://github.com/ChenMnZ/PrefixQuant. |
| title | PrefixQuant: Eliminating Outliers by Prefixed Tokens for Large Language Models Quantization |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2410.05265 |