适配微调后的MiniCPM-V-2_6，总是出现错误：Failed to get vocab size from tokenizer wrapper with exception_MindIE_昇腾论坛

# 适配微调后的MiniCPM-V-2_6，总是出现错误：Failed to get vocab size from tokenizer wrapper with exception_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-02102189746891411110-1-1.html

**生成时间**: 2025-08-27 10:33:57

---

## 问题描述

芯片310P3 mindle2.0.RC2-300I-Duo-py311-openeuler24.03-lts 因为多模态大模型只支持MiniCPM-V-2_6因此选择了它微调前的大模型可以适配成功正常运行 使用rtx3090微调后则适配失败出现如下错误 [2025-08-05 11:20:20.078] [16979] [281473793777632] [llm] [ERROR] [model_deploy_config.cpp:102] Failed to get vocab size from tokenizer wrapper with exception: ValueError: Get instance from Minicpmqwen2v2Config failed. Please check the input parameters model_path and kwargs. If the input parameters are valid and the required files exist in model_path, make sure the folder's owner has execute permission. Otherwise, please check the function stack for detailed exception information. At: /usr/local/Ascend/atb-models/atb_llm/models/base/model_utils.py(312): wrapper /usr/local/Ascend/atb-models/atb_llm/models/minicpm_qwen2_v2/router_minicpm_qwen2_v2.py(41): get_config /usr/local/Ascend/atb-models/atb_llm/models/base/router.py(109): config /usr/local/Ascend/atb-models/atb_llm/models/minicpm_qwen2_v2/router_minicpm_qwen2_v2.py(20): __post_init__ <string>(23): __init_...

## 完整内容

芯片310P3 mindle2.0.RC2-300I-Duo-py311-openeuler24.03-lts 因为多模态大模型只支持MiniCPM-V-2_6因此选择了它微调前的大模型可以适配成功正常运行 使用rtx3090微调后则适配失败出现如下错误 [2025-08-05 11:20:20.078] [16979] [281473793777632] [llm] [ERROR] [model_deploy_config.cpp:102] Failed to get vocab size from tokenizer wrapper with exception: ValueError: Get instance from Minicpmqwen2v2Config failed. Please check the input parameters model_path and kwargs. If the input parameters are valid and the required files exist in model_path, make sure the folder's owner has execute permission. Otherwise, please check the function stack for detailed exception information. At: /usr/local/Ascend/atb-models/atb_llm/models/base/model_utils.py(312): wrapper /usr/local/Ascend/atb-models/atb_llm/models/minicpm_qwen2_v2/router_minicpm_qwen2_v2.py(41): get_config /usr/local/Ascend/atb-models/atb_llm/models/base/router.py(109): config /usr/local/Ascend/atb-models/atb_llm/models/minicpm_qwen2_v2/router_minicpm_qwen2_v2.py(20): __post_init__ <string>(23): __init__ /usr/local/Ascend/atb-models/atb_llm/models/__init__.py(110): get_model /usr/local/Ascend/atb-models/atb_llm/runner/tokenizer_wrapper.py(18): __init__ /usr/local/lib/python3.11/site-packages/mindie_llm/modeling/model_wrapper/atb/atb_tokenizer_wrapper.py(8): __init__ /usr/local/lib/python3.11/site-packages/mindie_llm/modeling/model_wrapper/__init__.py(28): get_tokenizer_wrapper 目前已经升级transformer与微调环境的完全一致并且将trustRemoteCode设置为true了

---

## 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**: 昇腾, NPU, AI, MindIE
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。*