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 16, 04:30 UTC
[V1][Bugfix] Skip hashing empty or None mm_data (#11386)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Woosuk Kwon committed
Dec 21, 2024 at 14:22 UTC
dd2b5633dd5fb0ecb5fb7247351ebedc1d69d054
1 file changed
+4
vllm/v1/engine/mm_input_mapper.py
+4
@@ -180,6 +180,10 @@ class MMHasher:
180
return None
181
182
mm_data = prompt["multi_modal_data"]
183
+ if not mm_data:
184
+ # mm_data can be None or an empty dict.
185
+ return None
186
+
187
image_inputs = mm_data["image"]
188
189
return self.hash_images(image_inputs)