Quick Review: Norm Tweaking: High-performance Low-bit Quantization of Large Language Models
作者:XD / 发表: 2023年12月6日 23:44 / 更新: 2023年12月6日 23:52 / 科研学习 / 阅读量:740
Norm Tweaking: High-performance Low-bit Quantization of Large Language Models
- Paper: Norm Tweaking on arXiv
- Code: None available
- Organization: Meituan
Steps for Implementation:
- Generate Data: Prepare and preprocess the dataset suitable for training the model.
- GPTQ: Apply GPTQ method for optimizing the quantization precision of model parameters.
- Train LayerNorm Only: Focus on training the Layer Normalization component of the model for fine-tuning and optimization.
相关标签