arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2502.07820cs.ARcs.AI
收藏

用于IMC阵列的低秩压缩

Low-Rank Compression for IMC Arrays

Sungkyunkwan University(成均馆大学);College of Information and Communication Engineering, Sungkyunkwan University(成均馆大学信息与通信工程学院)

Kang Eun Jeon, Johnny Rhe, Jong Hwan Ko

AI总结:

针对IMC架构中传统剪枝方法的面积与能耗问题,提出结合移位复制核映射技术和分组低秩卷积的低秩压缩方法,实现了比现有剪枝技术高达2.5倍加速或20.9%精度提升。

摘要:

本研究针对内存计算(IMC)架构中的低秩模型压缩挑战展开。传统剪枝方法虽能有效减小模型尺寸,但需额外外围电路管理复杂数据流并缓解错位问题,导致面积和能耗开销增加。为规避这些缺点,我们提出利用低秩压缩技术,与剪枝不同,该技术可简化数据流并与IMC架构无缝集成。然而,低秩压缩存在两大挑战:i)IMC阵列利用率欠佳;ii)精度受损。为此,我们引入新颖方法:i)采用移位复制核(SDK)映射技术,利用空闲IMC列进行并行处理;ii)分组低秩卷积,缓解分解矩阵中的信息失衡。实验结果表明,所提方法相较现有剪枝技术实现了高达2.5倍的加速或20.9%的精度提升。

英文摘要:

In this study, we address the challenge of low-rank model compression in the context of in-memory computing (IMC) architectures. Traditional pruning approaches, while effective in model size reduction, necessitate additional peripheral circuitry to manage complex dataflows and mitigate dislocation issues, leading to increased area and energy overheads. To circumvent these drawbacks, we propose leveraging low-rank compression techniques, which, unlike pruning, streamline the dataflow and seamlessly integrate with IMC architectures. However, low-rank compression presents its own set of challenges, namely i) suboptimal IMC array utilization and ii) compromised accuracy. To address these issues, we introduce a novel approach i) employing shift and duplicate kernel (SDK) mapping technique, which exploits idle IMC columns for parallel processing, and ii) group low-rank convolution, which mitigates the information imbalance in the decomposed matrices. Our experimental results demonstrate that our proposed method achieves up to 2.5x speedup or +20.9% accuracy boost over existing pruning techniques.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑