Accessing state_dict() of models trained with previous pytorch version #42242
Description
🐛 Bug
Since updating to version 1.6, when trying to access model.state_dict() of a older pytorch trained models (where model is a subclass of torch.nn.module), I'm getting a ModuleAttributeError.
To Reproduce
Steps to reproduce the behavior:
torch.load("path/to/model", map_location='cpu')
print(model.state_dict())
where the model was trained with an older version of pytorch.
The output is
File "/home/oliviermt/miniconda3/envs/projections/lib/python3.7/site-packages/torch/nn/modules/module.py", line 900, in state_dict
module.state_dict(destination, prefix + name + '.', keep_vars=keep_vars)
File "/home/oliviermt/miniconda3/envs/projections/lib/python3.7/site-packages/torch/nn/modules/module.py", line 900, in state_dict
module.state_dict(destination, prefix + name + '.', keep_vars=keep_vars)
File "/home/oliviermt/miniconda3/envs/projections/lib/python3.7/site-packages/torch/nn/modules/module.py", line 897, in state_dict
self._save_to_state_dict(destination, prefix, keep_vars)
File "/home/oliviermt/miniconda3/envs/projections/lib/python3.7/site-packages/torch/nn/modules/module.py", line 860, in _save_to_state_dict
if buf is not None and name not in self._non_persistent_buffers_set:
File "/home/oliviermt/miniconda3/envs/projections/lib/python3.7/site-packages/torch/nn/modules/module.py", line 772, in __getattr__
type(self).__name__, name))
torch.nn.modules.module.ModuleAttributeError: 'GaussianSmearing' object has no attribute '_non_persistent_buffers_set'
I was able to make it work by adding the following lines after the the torch.load:
for mod in model.modules():
if not hasattr(mod, "_non_persistent_buffers_set"):
mod._non_persistent_buffers_set = set()
If it's expected behaviour, then ignore this issue, but I would guess it's not.
Thanks a lot!
Expected behavior
Not needing to add those 3 lines.
Environment
PyTorch version: 1.6.0
Is debug build: No
CUDA used to build PyTorch: 10.2
OS: Ubuntu 16.04.6 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
CMake version: version 3.5.1
Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
Versions of relevant libraries:
[pip3] numpy==1.19.1
[pip3] torch==1.6.0
[pip3] torchvision==0.7.0
[conda] numpy 1.19.1 pypi_0 pypi
[conda] torch 1.6.0 pypi_0 pypi
[conda] torchvision 0.7.0 pypi_0 pypi