TranSpa: Towards Efficient Structured Sparse Training for Transformers
Keywords: Sparse Training, Transformer, Efficient Inference, Efficient Training
Abstract: Transformers have emerged as the backbone neural network architecture in today's AI applications. Due to their high complexity, sparsifying transformers, at both pre-training and fine-tuning stages, is very attractive for lower the training and inference costs. In this paper, we propose TranSpa, an efficient structured sparse training approach for language and vision transformers. Unlike prior works focusing on individual building blocks, TranSpa fully considers the correlation between the weight matrices and their component rows/columns, and performs the coupled estimation and coupled sparsification. To achieve that, TranSpa introduces the use of new granularity when calibrating the importance of structural components in the transformer and removing the insignificant parts. Evaluations across different models, in both pre-training and fine-tuning scenarios, demonstrate the effectiveness of the proposed approach. TranSpa can bring $1.6\times$ size reduction with $0.6$ lower perplexity when training GPT-2 model from scratch. It also enables $1.6\times$ training speedup over the existing sparse pre-training method. For training sparse LLaMA-1B from scratch, our approach reduces GPU memory usage by 50\%, decreases training time by 21\%, and achieves a $1.6\times$ speedup in inference throughput while maintaining model performance. Experiments of applying TranSpa for fine-tuning tasks also show significant performance improvement with respect to model accuracy and pruning cost reduction.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 4743
Loading