pytorch中出现AttributeError: Can‘t get attribute ‘Net‘ on ＜module ‘main‘＞错误

最新推荐文章于 2024-07-18 10:25:43 发布

原创

最新推荐文章于 2024-07-18 10:25:43 发布 · 2.4k 阅读

2 ·

CC 4.0 BY-SA版权

文章标签：

#pytorch #深度学习 #python

本文解决了一个关于PyTorch模型加载时遇到的AttributeError问题。问题源于新环境中未能找到自定义的Net类，通过将Net类的定义包含在加载脚本中解决了该问题。

问题报错如下：
AttributeError                            Traceback (most recent call last)
Cell In[10], line 1
----> 1 model = torch.load('fruit30_pytorch_light.pth')
      2 model = model.eval().to(device)

File E:\Anaconda\envs\pytorch\lib\site-packages\torch\serialization.py:712, in load(f, map_location, pickle_module, **pickle_load_args)
    710             opened_file.seek(orig_position)
    711             return torch.jit.load(opened_file)
--> 712         return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
    713 return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)

File E:\Anaconda\envs\pytorch\lib\site-packages\torch\serialization.py:1049, in _load(zip_file, map_location, pickle_module, pickle_file, **pickle_load_args)
   1047 unpickler = UnpicklerWrapper(data_file, **pickle_load_args)
   1048 unpickler.persistent_load = persistent_load
-> 1049 result = unpickler.load()
   1051 torch._utils._validate_loaded_sparse_