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 12, 03:30 UTC
[BugFix] fix imgs_pos in hunyuan_vl (#29879)
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
JackieWu committed
Dec 3, 2025 at 14:20 UTC
0bec63fa317e1fbd62e19b0fc31c43c81bf89077
1 file changed
+1
-1
vllm/transformers_utils/processors/hunyuan_vl.py
+1
-1
@@ -123,7 +123,7 @@ class HunYuanVLProcessor(ProcessorMixin):
123
124
attention_mask = input_ids.ne(self.pad_id)
125
text_inputs["attention_mask"] = attention_mask
126
- text_inputs["imgs_pos"] = [self.get_imgs_pos(input_ids)]
126
+ text_inputs["imgs_pos"] = [self.get_imgs_pos(e) for e in input_ids]
127
# image_inputs["imgs"] = [[image_inputs["pixel_values"]]]
128
129
return_tensors = kwargs.pop("return_tensors", None)