[Core] Simplify the Dp padding/should ubatch coordination logic (#25768)

Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Sage Moore
2025-10-06 18:57:49 -07:00
committed by GitHub
parent c50901f3b9
commit 2111b4643c
10 changed files with 297 additions and 462 deletions

View File

@@ -152,6 +152,10 @@ class ParallelConfig:
threshold, microbatching will be used. Otherwise, the request will be
processed in a single batch."""
disable_nccl_for_dp_synchronization: bool = False
"""Forces the dp synchronization logic in vllm/v1/worker/dp_utils.py
to use Gloo instead of NCCL for its all reduce"""
ray_workers_use_nsight: bool = False
"""Whether to profile Ray workers with nsight, see https://docs.ray.io/en/latest/ray-observability/user-guides/profiling.html#profiling-nsight-profiler."""