From 0fbf42af841993ab1c189efca34de6b9799526b7 Mon Sep 17 00:00:00 2001 From: djramic Date: Sat, 20 Jun 2026 00:20:59 +0200 Subject: [PATCH] [ROCm] Fix VRAM not freed in test_phi3v (#46046) Signed-off-by: Djordje Ramic --- .buildkite/test_areas/models_multimodal.yaml | 1 - tests/models/multimodal/pooling/test_phi3v.py | 2 ++ 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/.buildkite/test_areas/models_multimodal.yaml b/.buildkite/test_areas/models_multimodal.yaml index a7358e8dbd..27e73e55a3 100644 --- a/.buildkite/test_areas/models_multimodal.yaml +++ b/.buildkite/test_areas/models_multimodal.yaml @@ -68,7 +68,6 @@ steps: - cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model # Otherwise, mp_method="spawn" doesn't work mirror: amd: - soft_fail: true device: mi325_1 depends_on: - image-build-amd diff --git a/tests/models/multimodal/pooling/test_phi3v.py b/tests/models/multimodal/pooling/test_phi3v.py index 285ded375d..ba017f065a 100644 --- a/tests/models/multimodal/pooling/test_phi3v.py +++ b/tests/models/multimodal/pooling/test_phi3v.py @@ -2,6 +2,7 @@ # SPDX-FileCopyrightText: Copyright contributors to the vLLM project import pytest +import torch import torch.nn.functional as F import transformers.utils from PIL import Image @@ -52,6 +53,7 @@ def _get_cherry_blossom_image() -> Image.Image: ) +@torch.inference_mode() def _run_test( hf_runner: type[HfRunner], vllm_runner: type[VllmRunner],