arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.07306cs.DCcs.LG
收藏

并行轨道变换器:通过减少同步实现快速GPU推理

Parallel Track Transformers: Enabling Fast GPU Inference with Reduced Synchronization

Apple(苹果公司)

Chong Wang, Nan Du, Tom Gunter, Tao Lei, Kulin Seth, Senyu Tong, Jianyu Wang, Guoli Yin, Xiyou Zhou, Kelvin Zou, Ruoming Pang

AI总结:

并行轨道变换器通过减少同步操作提升GPU推理效率,适用于大规模LLM服务。

摘要:

高效的大规模变压器大型语言模型(LLM)推理仍是一个基本的系统挑战,通常需要多GPU并行化来满足严格的延迟和吞吐量目标。传统张量并行将矩阵运算分解到设备上,但引入了显著的跨GPU同步,导致通信瓶颈和性能下降。我们提出了并行轨道(PT)变换器,一种新的架构范式,通过重新组织计算来最小化跨设备依赖性。PT相对于标准张量并行在同步操作上减少了高达16倍,同时在我们的实验中保持了具有竞争力的模型质量。我们将PT集成到两个广泛采用的LLM服务堆栈中-Tensor-RT-LLM和vLLM,并报告了一致的服务效率提升,包括首次令牌时间减少15-30%,每个输出令牌时间减少2-12%,以及在两种设置中吞吐量增加高达31.90%。

英文摘要:

Efficient large-scale inference of transformer-based large language models (LLMs) remains a fundamental systems challenge, frequently requiring multi-GPU parallelism to meet stringent latency and throughput targets. Conventional tensor parallelism decomposes matrix operations across devices but introduces substantial inter-GPU synchronization, leading to communication bottlenecks and degraded scalability. We propose the Parallel Track (PT) Transformer, a novel architectural paradigm that restructures computation to minimize cross-device dependencies. PT achieves up to a 16x reduction in synchronization operations relative to standard tensor parallelism, while maintaining competitive model quality in our experiments. We integrate PT into two widely adopted LLM serving stacks-Tensor-RT-LLM and vLLM-and report consistent improvements in serving efficiency, including up to 15-30% reduced time to first token, 2-12% reduced time per output token, and up to 31.90% increased throughput in both settings.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑