تقرير
FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
العنوان: | FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision |
---|---|
المؤلفون: | Shah, Jay, Bikshandi, Ganesh, Zhang, Ying, Thakkar, Vijay, Ramani, Pradeep, Dao, Tri |
سنة النشر: | 2024 |
المجموعة: | Computer Science |
مصطلحات موضوعية: | Computer Science - Machine Learning, Computer Science - Artificial Intelligence |
الوصف: | Attention, as a core layer of the ubiquitous Transformer architecture, is the bottleneck for large language models and long-context applications. FlashAttention elaborated an approach to speed up attention on GPUs through minimizing memory reads/writes. However, it has yet to take advantage of new capabilities present in recent hardware, with FlashAttention-2 achieving only 35% utilization on the H100 GPU. We develop three main techniques to speed up attention on Hopper GPUs: exploiting asynchrony of the Tensor Cores and TMA to (1) overlap overall computation and data movement via warp-specialization and (2) interleave block-wise matmul and softmax operations, and (3) block quantization and incoherent processing that leverages hardware support for FP8 low-precision. We demonstrate that our method, FlashAttention-3, achieves speedup on H100 GPUs by 1.5-2.0$\times$ with FP16 reaching up to 740 TFLOPs/s (75% utilization), and with FP8 reaching close to 1.2 PFLOPs/s. We validate that FP8 FlashAttention-3 achieves 2.6$\times$ lower numerical error than a baseline FP8 attention. |
نوع الوثيقة: | Working Paper |
URL الوصول: | http://arxiv.org/abs/2407.08608 |
رقم الأكسشن: | edsarx.2407.08608 |
قاعدة البيانات: | arXiv |
كن أول من يترك تعليقا!