Stop setting CUDA_VISIBLE_DEVICES internally in vLLM, add device_ids arg (#45026)

Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Codex <codex@openai.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: kourosh hakhamaneshi <kouroshHakha@users.noreply.github.com>
This commit is contained in:
Tyler Michael Smith
2026-06-20 13:38:10 -07:00
committed by GitHub
co-authored by Claude Codex mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> kourosh hakhamaneshi
parent e9de72fe6c
commit ebfbcfe46a
24 changed files with 722 additions and 185 deletions
@@ -704,7 +704,14 @@ class FlashInferNVLinkOneSidedManager(All2AllManagerBase):
self.num_experts = num_experts
self.cleanup()
gpus_per_node = torch.accelerator.device_count()
from vllm.platforms.interface import get_assigned_physical_gpu_ids
assigned_physical_gpu_ids = get_assigned_physical_gpu_ids()
gpus_per_node = (
len(assigned_physical_gpu_ids)
if assigned_physical_gpu_ids is not None
else torch.accelerator.device_count()
)
logger.debug(
"Making One-sided NVLink mapping: rank=%d, world size=%d",
self.rank,