From 21ee743f474a35a056e8d5ade2c9aa744e726e6e Mon Sep 17 00:00:00 2001 From: "Kevin H. Luu" Date: Wed, 23 Sep 2026 16:11:50 -0700 Subject: [PATCH] [CI][Bugfix] Limit MRV2 sampler JIT warmup registration to ROCm (#58465) Signed-off-by: khluu Co-authored-by: Claude Opus 5.5 --- vllm/v1/worker/gpu/model_runner.py | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/vllm/v1/worker/gpu/model_runner.py b/vllm/v1/worker/gpu/model_runner.py index 681b29fcb9..75c9c412cf 100644 --- a/vllm/v1/worker/gpu/model_runner.py +++ b/vllm/v1/worker/gpu/model_runner.py @@ -55,6 +55,7 @@ from vllm.multimodal import MULTIMODAL_REGISTRY from vllm.multimodal.encoder_budget import ( MultiModalBudget, ) +from vllm.platforms import current_platform from vllm.sequence import IntermediateTensors from vllm.tasks import SupportedTask from vllm.utils.gc_utils import freeze_gc_for_cudagraph_capture @@ -439,14 +440,15 @@ class GPUModelRunner(LoRAModelRunnerMixin): self.vllm_config, self.req_states, self.is_pooling_model, config_processors ) if self.is_last_pp_rank and not self.is_pooling_model: - from vllm.v1.sample.ops.topk_topp_sampler import ( - register_top_k_top_p_warmups, - ) - # V2 bypasses TopKTopPSampler, which registers native warmups. - # CUDA also needs these for its FlashInfer fallback paths. - with self.jit_warmup_registry.activate(): - register_top_k_top_p_warmups() + # ROCm only: on CUDA these warmups add ~2 min to every engine start. + if current_platform.is_rocm(): + from vllm.v1.sample.ops.topk_topp_sampler import ( + register_top_k_top_p_warmups, + ) + + with self.jit_warmup_registry.activate(): + register_top_k_top_p_warmups() sampler_kwargs: dict[str, Any] = { "vllm_config": self.vllm_config,