Error with some pretrained models: EOFError: Ran out of input when loading model

[PabloSczn]

Hello,

I'm encountering an issue when loading pretrained models from the OpenXAI repository. Specifically, when I attempt to generate explanations, I receive the following error with model files for various datasets with logistic regression:

Data: german, Model: lr
Traceback (most recent call last):
  File "...\generate_explanations.py", line 42, in <module>
    model = LoadModel(data_name, model_name, pretrained=pretrained)
  File "...\model.py", line 42, in LoadModel
    state_dict = torch.load(model_path+model_filename, map_location=torch.device('cpu'))
  File "...\torch\serialization.py", line 1114, in load
    return _legacy_load(
  File "...\torch\serialization.py", line 1338, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
EOFError: Ran out of input

It appears that the .pt model file being loaded is either incomplete or corrupted.
I tried downloading the model file directly from the Dataverse link provided:

• Download link: https://dataverse.harvard.edu/api/access/datafile/8550945

After downloading the .pt file from this link, I attempted to generate explanations again, but it failed once again