SmartMem: Layout Transformation Elimination and Adaptation for Efficient DNN Execution on Mobile
Authors:
- Wei NiuWNWei Niu
- Musfiqur Rahman SanimMSMusfiqur Rahman Sanim
- Zhihao ShuZSZhihao Shu
- Jiexiong GuanJGJiexiong Guan
- Xipeng ShenXSXipeng Shen
- Miao YinMYMiao Yin
- Gagan AgrawalGAGagan Agrawal
- Bin RenBRBin Ren
Preprint
Abstract
This work is motivated by recent developments in Deep Neural Networks, particularly the Transformer architectures underlying applications such as ChatGPT, and the need for performing inference on mobile devices. Focusing on emerging transformers (specifically the ones with computationally efficient Swin-like architectures) and large models (e.g., Stable Diffusion and LLMs) based on transformers, we observe that layout transformations between the computational operators cause a significant slowdown in these applications. This paper presents SmartMem, a comprehensive framework for eliminating most layout transformations, with the idea that multiple operators can use the same tensor layout through careful choice of layout and implementation of operations. Our approach is based on classifying the operators into four groups, and considering combinations of producer-consumer edges between the operators. We develop a set of methods for searching such layouts. Another component of our work is developing efficient memory layouts for 2.5 dimensional memory commonly seen in mobile devices. Our experimental results show that SmartMem outperforms 5 state-of-the-art DNN execution frameworks on mobile devices across 18 varied neural networks, including CNNs, Transformers with both local and global attention, as well as LLMs. In particular, compared to DNNFusion, SmartMem achieves an average speedup of 2.8$\times$, and outperforms TVM and MNN with speedups of 6.9$\times$ and 7.9$\times$, respectively, on average.Keywords
All Related Versions
- Version 1, 2024-04-21, ArXiv
- Published version: , 916.
This publication has 5 references indexed in Scilit:
- Scale-Aware Modulation Meet Transformer2023
- TensorIR: An Abstraction for Automatic Tensorized Program Optimization | Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 22023
- CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows2022
- LANet: Local Attention Embedding to Improve the Semantic Segmentation of Remote Sensing ImagesIEEE Transactions on Geoscience and Remote Sensing, 2020
- Fusion of loops for parallelism and localityIEEE Transactions on Parallel and Distributed Systems, 1997