[ROCm][CI] Explicitly tear down multimodal offline LLMs (#46999)

Signed-off-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
Andreas Karatzas
2026-06-29 07:59:24 +00:00
committed by GitHub
parent f6bb8682ee
commit db28ae2d07
6 changed files with 91 additions and 29 deletions
@@ -61,8 +61,9 @@ def test_mm_cache_stats(
image_urls,
mm_processor_cache_type,
caplog_vllm,
multimodal_llm_factory,
):
llm = LLM(
llm = multimodal_llm_factory(
model="llava-hf/llava-1.5-7b-hf",
max_model_len=4096,
max_num_seqs=5,