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 13, 08:15 UTC
[CI] Skip "Multi-Modal Models Test (Extended) 3" test that's broken in current Transformers (#28559)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Harry Mellor committed
Nov 12, 2025 at 19:38 UTC
a39dd7bb06c3bea055057d5c272ca952e0e000bf
1 file changed
+8
tests/models/multimodal/generation/test_common.py
+8
@@ -9,11 +9,13 @@ from collections import defaultdict
9
from pathlib import PosixPath
10
11
import pytest
12
+from packaging.version import Version
13
from transformers import (
14
AutoModel,
15
AutoModelForImageTextToText,
16
AutoModelForTextToWaveform,
17
)
18
+from transformers import __version__ as TRANSFORMERS_VERSION
19
20
from vllm.platforms import current_platform
21
from vllm.utils.func_utils import identity
@@ -851,6 +853,12 @@ VLM_TEST_SETTINGS = {
853
limit_mm_per_prompt={"image": 4},
854
)
855
],
856
+ marks=[
857
+ pytest.mark.skipif(
858
+ Version(TRANSFORMERS_VERSION) == Version("4.57.1"),
859
+ reason="This model is broken in Transformers v4.57.1",
860
+ )
861
+ ],
862
),
863
# regression test for https://github.com/vllm-project/vllm/issues/15122
864
"qwen2_5_vl-windows-attention": VLMTestInfo(