Bishop: Sparsified Bundling Spiking Transformers on Heterogeneous Cores with Error-Constrained Pruning

TL;DR

Bishop is a specialized hardware accelerator for spiking transformers that leverages spatiotemporal sparsity and error-constrained pruning to significantly improve speed and energy efficiency while maintaining high accuracy.

Contribution

The paper introduces Bishop, the first hardware architecture and co-design framework for spiking transformers that exploits spatiotemporal sparsity and error-bound pruning.

Findings

Achieves 5.91x speedup over previous accelerators.

Provides 6.11x energy efficiency improvement.

Delivers higher accuracy across multiple datasets.

Abstract

We present Bishop, the first dedicated hardware accelerator architecture and HW/SW co-design framework for spiking transformers that optimally represents, manages, and processes spike-based workloads while exploring spatiotemporal sparsity and data reuse. Specifically, we introduce the concept of Token-Time Bundle (TTB), a container that bundles spiking data of a set of tokens over multiple time points. Our heterogeneous accelerator architecture Bishop concurrently processes workload packed in TTBs and explores intra- and inter-bundle multiple-bit weight reuse to significantly reduce memory access. Bishop utilizes a stratifier, a dense core array, and a sparse core array to process MLP blocks and projection layers. The stratifier routes high-density spiking activation workload to the dense core and low-density counterpart to the sparse core, ensuring optimized processing tailored to the given spatiotemporal sparsity level. To further reduce data access and computation, we introduce a novel Bundle Sparsity-Aware (BSA) training pipeline that enhances not only the overall but also structured TTB-level firing sparsity. Moreover, the processing efficiency of self-attention layers is boosted by the proposed Error-Constrained TTB Pruning (ECP), which trims activities in spiking queries, keys, and values both before and after the computation of spiking attention maps with a well-defined error bound. Finally, we design a reconfigurable TTB spiking attention core to efficiently compute spiking attention maps by executing highly simplified "AND" and "Accumulate" operations. On average, Bishop achieves a 5.91x speedup and 6.11x improvement in energy efficiency over previous SNN accelerators, while delivering higher accuracy across multiple datasets.