MindIE版本：2.0.RC2

CANN版本：8.1.RC1

操作步骤：拉取Atlas800IA2版MindIE镜像，进入容器安装910A版toolkit、kernel、nnal。修改完镜像后，mindie-service拉起Qwen2.5-VL-7B成功，可以进行纯文本推理，但进行图像推理报错。

报错信息：

[ERROR] standard_model.py:194 - [Model] >>>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 "/root/anaconda3/envs/python311/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 183, in execute 

    return self._prefill(requests) 

           ^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 217, in _prefill 

    return self.__generate(requests, is_prefill=True, is_mix=False) 

           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 348, in __generate 

    request_tokens_np, eos_np, truncation_indices = self.__handle_requests(requests, is_prefill, is_mix) 

                                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/model_wrapper/standard_model.py", line 532, in __handle_requests 

    request_tokens_np, eos_np, _, truncation_indices = self.generator.generate_token(metadata) 

                                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/text_generator/generator.py", line 181, in generate_token 

    raise e 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/text_generator/generator.py", line 157, in generate_token 

    self.plugin.generate_token(input_metadata)) 

    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/utils/decorators/time_decorator.py", line 38, in wrapper 

    return func(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/text_generator/plugins/plugin.py", line 17, in generate_token 

    logits = self.generator_backend.forward(model_inputs) 

             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/utils/decorators/time_decorator.py", line 38, in wrapper 

    return func(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/text_generator/adapter/generator_torch.py", line 153, in forward 

    logits = self.model_wrapper.forward(model_inputs, self.cache_pool.npu_cache, **kwargs) 

             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/modeling/model_wrapper/atb/atb_model_wrapper.py", line 89, in forward 

    logits = self.forward_tensor( 

             ^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/mindie_llm/modeling/model_wrapper/atb/atb_model_wrapper.py", line 116, in forward_tensor 

    logits = self.model_runner.forward( 

             ^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/usr/local/Ascend/atb-models/atb_llm/runner/model_runner.py", line 297, in forward 

    res = self.model.forward(**kwargs) 

          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/usr/local/Ascend/atb-models/atb_llm/models/qwen2_vl/flash_causal_qwen2_vl.py", line 118, in forward 

    inputs_embeds, image_grid_thw, video_grid_thw, second_per_grid_ts = self.prepare_prefill_token_service( 

                                                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/usr/local/Ascend/atb-models/atb_llm/models/qwen2_vl/flash_causal_qwen2_vl.py", line 197, in prepare_prefill_token_service 

    image_features = self.vision_tower(image_pixel.to(self.vision_tower.dtype), image_grid_thw) 

                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl 

    return self._call_impl(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl 

    return forward_call(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/usr/local/Ascend/atb-models/atb_llm/models/qwen2_vl/modeling_qwen2_5_vl_vit_atb.py", line 867, in forward 

    vision_features = self.encoder( 

                      ^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl 

    return self._call_impl(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/root/anaconda3/envs/python311/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl 

    return forward_call(*args, **kwargs) 

           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

  File "/usr/local/Ascend/atb-models/atb_llm/models/qwen2_vl/modeling_qwen2_5_vl_vit_atb.py", line 714, in forward 

    hidden_states = self.graph.forward(self.graph_inputs, self.graph_outputs, self.graph_param) 

                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 

RuntimeError: Qwen25VL_VIT_graph nodes[1] infershape  fail, enable log: export ASDOPS_LOG_LEVEL=ERROR, export ASDOPS_LOG_TO_STDOUT=1