Optimized Speculative Sampling for GPU Hardware Accelerators

Dominik Wagner, Seanie Lee, Ilja Baumann, Philipp Seeberger, Korbinian Riedhammer, Tobias Bocklet


Abstract
In this work, we optimize speculative sampling for parallel hardware accelerators to improve sampling speed. We notice that substantial portions of the intermediate matrices necessary for speculative sampling can be computed concurrently. This allows us to distribute the workload across multiple GPU threads, enabling simultaneous operations on matrix segments within thread blocks. This results in profiling time improvements ranging from 6% to 13% relative to the baseline implementation, without compromising accuracy. To further accelerate speculative sampling, probability distributions parameterized by softmax are approximated by sigmoid. This approximation approach results in significantly greater relative improvements in profiling time, ranging from 37% to 94%, with a minor decline in accuracy. We conduct extensive experiments on both automatic speech recognition and summarization tasks to validate the effectiveness of our optimization methods.
Anthology ID:
2024.emnlp-main.370
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
6442–6458
Language:
URL:
https://aclanthology.org/2024.emnlp-main.370
DOI:
10.18653/v1/2024.emnlp-main.370
Bibkey:
Cite (ACL):
Dominik Wagner, Seanie Lee, Ilja Baumann, Philipp Seeberger, Korbinian Riedhammer, and Tobias Bocklet. 2024. Optimized Speculative Sampling for GPU Hardware Accelerators. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 6442–6458, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Optimized Speculative Sampling for GPU Hardware Accelerators (Wagner et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.370.pdf
Software:
 2024.emnlp-main.370.software.zip