Neural networks (NNs) are increasingly employed in safety-critical domains and in environments prone to unreliability (e.g., soft errors), such as on spacecraft. Therefore, it is critical to impart fault tolerance to NN inference. Algorithm-based fault tolerance (ABFT) is emerging as an efficient approach for fault tolerance in NNs. We propose an adaptive approach to ABFT for NN inference that exploits untapped opportunities in emerging deployment scenarios. GPUs have high compute-to-memory-bandwidth ratios, while NN layers have a wide range of arithmetic intensities. This leaves some layers compute bound and others memory-bandwidth bound, but current approaches to ABFT do not consider these differences. We first investigate ABFT schemes best suited for each of these scenarios. We then propose intensity-guided ABFT, an adaptive, arithmetic-intensity-guided approach that selects the most efficient ABFT scheme for each NN layer. Intensity-guided ABFT reduces execution-time overhead by 1.09--5.3$\times$ across many NNs compared to traditional approaches to ABFT.
翻译:神经网络(NNS)越来越多地用于安全关键领域和不可靠的环境(如软差错),如航天器,因此,对 NN 的推论给予过失容忍度至关重要。基于Agorithm 的过失容忍度(ABFT)正在形成,作为NNS的过失容忍度的一种有效办法。我们提议对ABFT的无过失推断采用适应性办法,利用新兴部署情景中尚未利用的机会。GPU具有高计算至模拟-带宽比率,而NNT层具有广泛的算术强度。这样,有些层的内装束和其他内存-带宽宽度的内存容忍度,但目前对ABFT的做法并不考虑这些差异。我们首先调查ABFT最适合其中每一种情况的办法。然后我们提出以强度为指南的ABFT,一种适应性、算术-密集性指导办法,为每一层选择最有效率的ABFT计划。NFT的强度由10.9-5.3 将执行时间比A-NFT的1.09-5.3-NFM0.xxxxxx 1.09-0.0.0.0.5-NFTxxxxxxxxxxxxxxxxxxxxx0.0.0.0.0.0.0.0.0.0.0.0.0.0.1-NFTFTFTFT_xxxxxxxxxxxxxxxxxxxx0.0.0.0.0.0.0.3xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx。