LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design
This work addresses the problem of efficient inference for Mamba models, which are important for AI applications requiring long sequences, but it is incremental as it focuses on hardware optimization rather than algorithmic breakthroughs.
The paper tackles the challenge of accelerating Mamba state space models, which are inefficient on existing hardware due to activation outliers and complex dependencies, by proposing LightMamba, a co-design of quantization and FPGA accelerator architecture that achieves up to 6.06x higher energy efficiency and 1.43x faster token generation compared to GPU baselines.
State space models (SSMs) like Mamba have recently attracted much attention. Compared to Transformer-based large language models (LLMs), Mamba achieves linear computation complexity with the sequence length and demonstrates superior performance. However, Mamba is hard to accelerate due to the scattered activation outliers and the complex computation dependency, rendering existing LLM accelerators inefficient. In this paper, we propose LightMamba that co-designs the quantization algorithm and FPGA accelerator architecture for efficient Mamba inference. We first propose an FPGA-friendly post-training quantization algorithm that features rotation-assisted quantization and power-of-two SSM quantization to reduce the majority of computation to 4-bit. We further design an FPGA accelerator that partially unrolls the Mamba computation to balance the efficiency and hardware costs. Through computation reordering as well as fine-grained tiling and fusion, the hardware utilization and memory efficiency of the accelerator get drastically improved. We implement LightMamba on Xilinx Versal VCK190 FPGA and achieve 4.65x to 6.06x higher energy efficiency over the GPU baseline. When evaluated on Alveo U280 FPGA, LightMamba reaches 93 tokens/s, which is 1.43x that of the GPU baseline. Our code is available at https://github.com/PKU-SEC-Lab/LightMamba.