Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2506.16500 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912441996148736 |
|---|---|
| author | Khaki, Samir Li, Xiuyu Guo, Junxian Zhu, Ligeng Xu, Chenfeng Plataniotis, Konstantinos N. Yazdanbakhsh, Amir Keutzer, Kurt Han, Song Liu, Zhijian |
| author_facet | Khaki, Samir Li, Xiuyu Guo, Junxian Zhu, Ligeng Xu, Chenfeng Plataniotis, Konstantinos N. Yazdanbakhsh, Amir Keutzer, Kurt Han, Song Liu, Zhijian |
| contents | Fine-tuning LLMs is both computationally and memory-intensive. While parameter-efficient fine-tuning methods, such as QLoRA and DoRA, reduce the number of trainable parameters and lower memory usage, they do not decrease computational cost. In some cases, they may even slow down fine-tuning. In this paper, we introduce SparseLoRA, a method that accelerates LLM fine-tuning through contextual sparsity. We propose a lightweight, training-free SVD sparsity estimator that dynamically selects a sparse subset of weights for loss and gradient computation. Also, we systematically analyze and address sensitivity across layers, tokens, and training steps. Our experimental results show that SparseLoRA reduces computational cost by up to 2.2 times and a measured speedup of up to 1.6 times while maintaining accuracy across various downstream tasks, including commonsense and arithmetic reasoning, code generation, and instruction following. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_16500 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity Khaki, Samir Li, Xiuyu Guo, Junxian Zhu, Ligeng Xu, Chenfeng Plataniotis, Konstantinos N. Yazdanbakhsh, Amir Keutzer, Kurt Han, Song Liu, Zhijian Machine Learning Fine-tuning LLMs is both computationally and memory-intensive. While parameter-efficient fine-tuning methods, such as QLoRA and DoRA, reduce the number of trainable parameters and lower memory usage, they do not decrease computational cost. In some cases, they may even slow down fine-tuning. In this paper, we introduce SparseLoRA, a method that accelerates LLM fine-tuning through contextual sparsity. We propose a lightweight, training-free SVD sparsity estimator that dynamically selects a sparse subset of weights for loss and gradient computation. Also, we systematically analyze and address sensitivity across layers, tokens, and training steps. Our experimental results show that SparseLoRA reduces computational cost by up to 2.2 times and a measured speedup of up to 1.6 times while maintaining accuracy across various downstream tasks, including commonsense and arithmetic reasoning, code generation, and instruction following. |
| title | SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2506.16500 |