A technical paper titled “A Comprehensive Performance Study of Large Language Models on Novel AI Accelerators” was published by researchers at Argonne National Laboratory, State University of New York, and University of Illinois.
“Artificial intelligence (AI) methods have become critical in scientific applications to help accelerate scientific discovery. Large language models (LLMs) are being considered as a promising approach to address some of the challenging problems because of their superior generalization capabilities across domains. The effectiveness of the models and the accuracy of the applications is contingent upon their efficient execution on the underlying hardware infrastructure. Specialized AI accelerator hardware systems have recently become available for accelerating AI applications. However, the comparative performance of these AI accelerators on large language models has not been previously studied. In this paper, we systematically study LLMs on multiple AI accelerators and GPUs and evaluate their performance characteristics for these models. We evaluate these systems with (i) a micro-benchmark using a core transformer block, (ii) a GPT- 2 model, and (iii) an LLM-driven science use case, GenSLM. We present our findings and analyses of the models’ performance to better understand the intrinsic capabilities of AI accelerators. Furthermore, our analysis takes into account key factors such as sequence lengths, scaling behavior, sparsity, and sensitivity to gradient accumulation steps.”
Find the technical paper here. Published October 2023 (preprint).
Emani, Murali, Sam Foreman, Varuni Sastry, Zhen Xie, Siddhisanket Raskar, William Arnold, Rajeev Thakur, Venkatram Vishwanath, and Michael E. Papka. “A Comprehensive Performance Study of Large Language Models on Novel AI Accelerators.” arXiv preprint arXiv:2310.04607 (2023).
Related Reading
Processor Tradeoffs For AI Workloads
Gaps are widening between technology advances and demands, and closing them is becoming more difficult.
How Much AI Is Really Needed?
Performance depends on the application it is being applied to.
Leave a Reply