基于MindIE2.0.RC2在300I Duo上单卡运行qwen2.5-vl-7b报错_MindIE_昇腾论坛

# 基于MindIE2.0.RC2在300I Duo上单卡运行qwen2.5-vl-7b报错_MindIE_昇腾论坛

## 概述

本文档基于昇腾社区论坛帖子生成的技术教程。

**原始链接**: https://www.hiascend.com/forum/thread-02116187962746668056-1-1.html

**生成时间**: 2025-08-27 10:33:58

---

## 问题描述

The old environment variable ATB_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable ATB_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable MINDIE_LLM_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable MINDIE_LLM_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_PATH will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_PATH as soon as possible. The ol...

## 完整内容

The old environment variable ATB_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable ATB_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable MINDIE_LLM_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable MINDIE_LLM_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_PATH will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_PATH as soon as possible. The old environment variable ATB_LOG_TO_FILE will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_FILE as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable MINDIE_LLM_PYTHON_LOG_TO_FILE will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_FILE as soon as possible. The old environment variable MINDIE_LLM_LOG_TO_FILE will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_FILE as soon as possible. The old environment variable OCK_LOG_LEVEL will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_LEVEL as soon as possible. The old environment variable OCK_LOG_TO_STDOUT will be deprecated on 2025/12/31. Please use the new environment variable MINDIE_LOG_TO_STDOUT as soon as possible. You are using a model of type qwen2_5_vl to instantiate a model of type . This is not supported for all configurations of models and can yield errors. Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.52, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`. You are using a model of type qwen2_5_vl to instantiate a model of type . This is not supported for all configurations of models and can yield errors. [2025-07-08 17:04:57.178431] [error] [1256320] [linear_operation.cpp:629] LinearOperation Error code: 8. value of inTensor0 k and value of inTensor1 k should be equal, inTensor0 k = 3424, inTensor1 k = 3420. Please check the shape of inTensors. [2025-07-08 17:04:57.178503] [error] [1256320] [operation_base.cpp:267] LinearOperation infer shape check fail, error code: 8 [2025-07-08 17:04:57.178515] [error] [1256320] [graph_operation.cpp:428] Qwen25VL_VIT_layer_0_graph node[12] infer shape fail, error code: 8 [2025-07-08 17:04:57.178566] [error] [1256320] [operation_base.cpp:273] Qwen25VL_VIT_layer_0_graph infer shape fail, error code: 8 Traceback (most recent call last): File "<frozen runpy>", line 198, in _run_module_as_main File "<frozen runpy>", line 88, in _run_code File "/usr/local/Ascend/atb-models/examples/models/qwen2_vl/run_pa.py", line 343, in <module> pa_runner.warm_up() File "/usr/local/Ascend/atb-models/examples/models/qwen2_vl/run_pa.py", line 141, in warm_up generate_req([single_req], self.model, self.max_batch_size, self.max_prefill_tokens, self.cache_manager) File "/usr/local/Ascend/atb-models/examples/server/generate.py", line 1143, in generate_req generate_token_with_clocking(model, cache_manager, batch) File "/usr/local/Ascend/atb-models/examples/server/generate.py", line 810, in generate_token_with_clocking res = generate_token(model, cache_manager, input_batch_in)  File "/usr/local/Ascend/atb-models/examples/server/generate.py", line 587, in generate_token logits = model.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 "/usr/local/lib64/python3.11/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl return self._call_impl(*args, **kwargs)  File "/usr/local/lib64/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_vi t_atb.py", line 867, in forward vision_features = self.encoder(  File "/usr/local/lib64/python3.11/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl return self._call_impl(*args, **kwargs)  File "/usr/local/lib64/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_vi t_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 [ERROR] 2025-07-08-17:05:02 (PID:1256320, Device:0, RankID:-1) ERR99999 UNKNOWN application exception 您好请问下您使用的配套的CANN版本和HDK版本是什么 使用的是300I-Duo-MindIE2.0.RC2镜像,HDK是24.1.rc2,cann是8.1.RC1已经解决了 需要修改/usr/local/Ascend/atb-models/atb_llm/models/qwen2_vl/modeling_qwen2_5_vl_vit_atb.py中class visionMlp(BaseMLP)中get_weight函数552行到560行如下 down_weight = down_weight_dict[f"{prefix}.down_proj.weight"] self.soc_info = NPUSocInfo() # tianzhang3 if self.soc_info.need_nz: # tianzhang3 down_weight = down_weight.transpose(0,1).contiguous() # tianzhang3 tp_intermediate_size = down_weight.shape[0] remainder = tp_intermediate_size % 32 if remainder != 0: tp_intermediate_size_pad = tp_intermediate_size + (32 - remainder) else: tp_intermediate_size_pad = tp_intermediate_size down_weight_padded = self.pad_tensor(down_weight, tp_intermediate_size_pad) if self.soc_info.need_nz: # tianzhang3 down_weight_padded = down_weight_padded.transpose(0,1).contiguous() # tianzhang3 down_weight_dict[f"{prefix}.down_proj.weight"] = down_weight_padded 似乎还是不行这个错误是什么呀 vision模型的中间层大小是3420实际送进去的是3424有问题

---

## 技术要点总结

基于以上内容，主要技术要点包括：

1. **问题类型**: 错误处理
2. **涉及技术**: NPU, MindIE, 昇腾, CANN, AI
3. **解决方案**: 请参考完整内容中的解决方案

## 相关资源

- 昇腾社区: https://www.hiascend.com/
- 昇腾论坛: https://www.hiascend.com/forum/

---

*本文档由AI自动生成，仅供参考。如有疑问，请参考原始帖子。*