BASS: Batched Attention-optimized Speculative Sampling

• Published in: arXiv.org 2024-06
• Main Authors: Qian, Haifeng, Gonugondla, Sujan Kumar, Ha, Sungsoo, Shang, Mingyue, Gouda, Sanjay Krishna, Nallapati, Ramesh, Sengupta, Sudipta, Ma, Xiaofei, Deoras, Anoop
• Format: Article
• Language: English
• Subjects: Budgets; Generative artificial intelligence; Large language models; Sequences
• ISSN: 2331-8422

Speculative decoding has emerged as a powerful method to improve latency and throughput in hosting large language models. However, most existing implementations focus on generating a single sequence. Real-world generative AI applications often require multiple responses and how to perform speculative decoding in a batched setting while preserving its latency benefits poses non-trivial challenges. This paper describes a system of batched speculative decoding that sets a new state of the art in multi-sequence generation latency and that demonstrates superior GPU utilization as well as quality of generations within a time budget. For example, for a 7.8B-size model on a single A100 GPU and with a batch size of 8, each sequence is generated at an average speed of 5.8ms per token, the overall throughput being 1.1K tokens per second. These results represent state-of-the-art latency and a 2.15X speed-up over optimized regular decoding. Within a time budget that regular decoding does not finish, our system is able to generate sequences with HumanEval Pass@First of 43% and Pass@All of 61%, far exceeding what's feasible with single-sequence speculative decoding. Our peak GPU utilization during decoding reaches as high as 15.8%, more than 3X the highest of that of regular decoding and around 10X of single-sequence speculative decoding.