Model loading error

[MingxinDing]

Is there an existing issue for this?

• I have searched the existing issues

Bug description

New commit failed to load model files.
cebra_error.md

Operating System

Windows 11

CEBRA version

0.4.0

Device type

gpu

Steps To Reproduce

No response

Relevant log output

/usr/local/lib/python3.10/dist-packages/cebra/integrations/sklearn/cebra.py:1436: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
  checkpoint = torch.load(filename, **kwargs)
/usr/local/lib/python3.10/dist-packages/sklearn/utils/_tags.py:354: FutureWarning: The CEBRA or classes from which it inherits use `_get_tags` and `_more_tags`. Please define the `__sklearn_tags__` method, or inherit from `sklearn.base.BaseEstimator` and/or other appropriate mixins such as `sklearn.base.TransformerMixin`, `sklearn.base.ClassifierMixin`, `sklearn.base.RegressorMixin`, and `sklearn.base.OutlierMixin`. From scikit-learn 1.7, not defining `__sklearn_tags__` will raise an error.
  warnings.warn(

---------------------------------------------------------------------------

AttributeError                            Traceback (most recent call last)

/usr/local/lib/python3.10/dist-packages/IPython/core/formatters.py in __call__(self, obj, include, exclude)
    968 
    969             if method is not None:
--> 970                 return method(include=include, exclude=exclude)
    971             return None
    972         else:

4 frames

/usr/local/lib/python3.10/dist-packages/sklearn/base.py in __sklearn_tags__(self)
    857 
    858     def __sklearn_tags__(self):
--> 859         tags = super().__sklearn_tags__()
    860         tags.transformer_tags = TransformerTags()
    861         return tags

AttributeError: 'super' object has no attribute '__sklearn_tags__'

---------------------------------------------------------------------------

AttributeError                            Traceback (most recent call last)

/usr/local/lib/python3.10/dist-packages/IPython/core/formatters.py in __call__(self, obj)
    343             method = get_real_method(obj, self.print_method)
    344             if method is not None:
--> 345                 return method()
    346             return None
    347         else:

4 frames

/usr/local/lib/python3.10/dist-packages/sklearn/base.py in __sklearn_tags__(self)
    857 
    858     def __sklearn_tags__(self):
--> 859         tags = super().__sklearn_tags__()
    860         tags.transformer_tags = TransformerTags()
    861         return tags

AttributeError: 'super' object has no attribute '__sklearn_tags__'

CEBRA(batch_size=2048, conditional='time_delta', learning_rate=0.0005,
      max_iterations=12000, model_architecture='offset10-model',
      num_hidden_units=64, output_dimension=3, temperature=1, time_offsets=10,
      verbose=True)

Anything else?

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[stes]

Thanks for flagging!

To understand the issue, you saved the model using a previous cebra version, and try to load it with the latest commit?

I'll attempt to repro the issue on my end.

[MingxinDing]

Yes I did originally. Then I saved models using the latest commit, same error with loading those as well.
Thanks very much!

[MMathisLab]

Which version of pytorch do you have installed? Thanks!

[MingxinDing]

2.5.1+cu121

[stes]

The error most likely occurred in #205 where the general logic behind the CEBRA model changed a bit. We had to adapt this due to some API changes in sklearn. I will see if we can ensure compatibility of model loading and keep you updated here.

In general, thanks for catching this, important issue to fix to keep the loading function backward compatible!

[stes]

@MingxinDing I could not reproduce your error yet.

Could you let me know which version of scikit-learn you are running?

I tried with 1.4x, 1.5x and 1.6.1 and was not able to repro the loading error, I was able to load a model saved in cebra 0.4.0 without issues.

Which version of CEBRA was used to save the model, and which version of CEBRA was used to load it?

[stes]

@MingxinDing could you check if you can load the following models in your environment?

• https://cebra.fra1.digitaloceanspaces.com/cebra_model_cebra-rc-scikit-learn-1.4.pt
• https://cebra.fra1.digitaloceanspaces.com/cebra_model_cebra-rc-scikit-learn-1.6.pt
• https://cebra.fra1.digitaloceanspaces.com/cebra_model_cebra-0.4.0-scikit-learn-1.6.pt
• https://cebra.fra1.digitaloceanspaces.com/cebra_model_cebra-0.4.0-scikit-learn-1.4.pt

That would be very valuable for debugging further. Thanks a lot!

[stes]

Hi @MingxinDing , if you get the chance, could you let me know if you have any additional insights on how to repro this issue?