#VPP

1 stories · 0 in 7 days · 0 in 30 days

Stories, products and papers tagged "VPP", newest first.

8月28日

PapersSelected73°17:39

VPP significantly improves efficiency for long-context LLM inference, reducing pipeline bubbles by 98% compared to DCPP on 512K-token sequences.

arXiv: DeepSeek@Yan Shi et al. (10)Source