A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Chore] Use `max_transformers_version` for Qwen-VL test (#26792)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Oct 14, 2025 at 18:03 UTC
d1d063a588205161fe5613d5a6d76311ce0dc83e
2 files changed
+2
-2
tests/models/multimodal/generation/test_common.py
-2
@@ -707,8 +707,6 @@ VLM_TEST_SETTINGS = {
707
max_num_seqs=2,
708
vllm_output_post_proc=model_utils.qwen_vllm_to_hf_output,
709
prompt_path_encoder=model_utils.qwen_prompt_path_encoder,
710
- # FIXME: https://github.com/huggingface/transformers/issues/38358
711
- marks=[pytest.mark.skip("Model initialization fails")],
710
),
711
"qwen2_vl": VLMTestInfo(
712
models=["Qwen/Qwen2-VL-2B-Instruct"],
tests/models/registry.py
+2
@@ -752,6 +752,8 @@ _MULTIMODAL_EXAMPLE_MODELS = {
752
"Qwen/Qwen-VL",
753
extras={"chat": "Qwen/Qwen-VL-Chat"},
754
trust_remote_code=True,
755
+ max_transformers_version="4.53.3",
756
+ transformers_version_reason="Use of deprecated imports which have been removed.", # noqa: E501
757
hf_overrides={"architectures": ["QwenVLForConditionalGeneration"]},
758
),
759
"Qwen2AudioForConditionalGeneration": _HfExamplesInfo(