Isn't possible to use IgFold anymore. Security vulnerability

[o-mathias]

When instaling and executing IgFold the error "ValueError: Due to a serious vulnerability issue in torch.load, even with weights_only=True, we now require users to upgrade torch to at least v2.6 in order to use the function. This version restriction does not apply when loading files with safetensors." is returned. I tried to use a previous version os torch, didn't worked. With the newest version of torch also didn't worked.

[xgqaikgh]

I used to meet a similar issue, because：
PyTorch has model files (such as .pt format), which contain not only parameters but also the model architecture. This means that when loading the model next time, you don’t need to configure the parameters or define the structure again. However, this also has a drawback: during loading, it may be necessary to execute code within the model, and the security of this code cannot be guaranteed.

To address this issue, PyTorch introduced the weights_onlyparameter. When this parameter is set to True, it means trusting the data in the model but not the code. When set to False, it means trusting both the data and the code in the model. In PyTorch < 2.6, the default value is False, while in PyTorch >= 2.6, the default is True.

The workaround is as follows. Add the following code at the beginning of your script:

import torch

# Temporarily modify the default weights_only behavior
original_load = torch.load
def custom_load(*args, **kwargs):
    if 'weights_only' not in kwargs:
        kwargs['weights_only'] = False
    return original_load(*args, **kwargs)
torch.load = custom_load

This method worked for me, though I'm not sure if it will work for you. There’s no need to change the PyTorch version.

[o-mathias]

Thank you so much @xgqaikgh ! Your solution worked very well and saved my pipeline!