环境信息：

• docker镜像：2.0.RC2-300I-Duo-py311-openeuler24.03-lts

• 硬件：4*300I DUO

问题现象：

在使用mindie-service启动qwen2.5-vl-3B后文字推理问答没问题，但是一旦进行图片理解推理就报错了，我尝试用Qwen2-VL-7B模型推理图片理解是没问题的

报错信息：

[2025-06-09 16:47:19.009+0800] [179] [281473458303328] [batchscheduler] [ERROR] [standard_model.py:188] : [Model] >>> global rank-2 Execute type:1, Exception:Qwen25VL_VIT_graph nodes[1] infershape fail, enable log: export ASDOPS_LOG_LEVEL=ERROR, export ASDOPS_LOG_TO_STDOUT=1 Traceback (most recent call last): File "/usr/local/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 176, in execute return self._prefill(requests) ^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 403, in _prefill return self.__generate(requests, is_prefill=True, is_mix=False) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 555, in __generate generate_output = self.__handle_requests(requests, is_prefill, is_mix) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 910, in __handle_requests

求助

这个是哪配置的不对还是驱动版本不对呀？有遇到过类似问题的小伙伴吗