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 9, 14:45 UTC
[CI] Fix MHA attention test failure (AttributeError when model_config is None in ViT attention backend) (#33033)
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Lucas Wilkinson committed
Jan 25, 2026 at 20:49 UTC
566cdb6cfb89d20faebe85f8796c3b03c38358e4
1 file changed
+4
-2
vllm/model_executor/models/vision.py
+4
-2
@@ -104,8 +104,9 @@ def get_vit_attn_backend(
104
"""
105
try:
106
vllm_config: VllmConfig = get_current_vllm_config()
107
+ model_config = vllm_config.model_config
108
multimodal_config: MultiModalConfig | None = (
108
- vllm_config.model_config.multimodal_config
109
+ model_config.multimodal_config if model_config is not None else None
110
)
111
except AssertionError:
112
multimodal_config = None
@@ -129,8 +130,9 @@ def is_vit_use_data_parallel():
130
"""
131
try:
132
vllm_config: VllmConfig = get_current_vllm_config()
133
+ model_config = vllm_config.model_config
134
multimodal_config: MultiModalConfig | None = (
133
- vllm_config.model_config.multimodal_config
135
+ model_config.multimodal_config if model_config is not None else None
136
)
137
except AssertionError:
138
multimodal_config = None