00:00
2026-06-24
rocm.blogs.amd.com
machine-learning
DP Attention and TBO for DeepSeek-V4 on MI355X
AMD introduces DP Attention and Two-Batch Overlap (TBO) optimizations for DeepSeek-V4 inference on MI355X GPUs, using a coordinated prefill scheduler called PrefillDelayer to reduce padding waste and โฆ