Issues loading ConvLSTM fire model

[RiccardoGrigoletto]

I am trying the notebooks on the Hopsworks platform.

In the second notebook (uc3-public-notebooks-main/2_UC3_DL_models_inference.ipynb), when I run convlstm = ConvLSTM_fire_model(**best_settings[model]).load_from_checkpoint(path_convlstm) it throws an error.

Here is the traceback.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-34-e7bd68891186> in <module>
      2 
      3 model = 'convlstm'
----> 4 convlstm = ConvLSTM_fire_model(**best_settings[model]).load_from_checkpoint(path_convlstm)

/srv/hops/staging/private_dirs/3df99fe2aee2df02abf1fd5010342a35f5b5e36320d0a6ba7b2033f4138852b3/localized_deps/greece_fire_models.py in __init__(self, dynamic_features, static_features, hidden_size, lstm_layers, lr, positive_weight, lr_scheduler_step, lr_scheduler_gamma, weight_decay)
     38         self.save_hyperparameters()
     39 
---> 40         self.model = SimpleConvLSTM(hparams=self.hparams)
     41 
     42         # loss function

/srv/hops/staging/private_dirs/3df99fe2aee2df02abf1fd5010342a35f5b5e36320d0a6ba7b2033f4138852b3/localized_deps/fire_modules.py in __init__(self, hparams)
     61                                  True,
     62                                  True,
---> 63                                  False)
     64 
     65         kernel_size = 3

/srv/hops/staging/private_dirs/3df99fe2aee2df02abf1fd5010342a35f5b5e36320d0a6ba7b2033f4138852b3/localized_deps/convlstm.py in __init__(self, input_dim, hidden_dim, kernel_size, num_layers, batch_first, bias, return_all_layers)
     91     def __init__(self, input_dim, hidden_dim, kernel_size, num_layers,
     92                  batch_first=False, bias=True, return_all_layers=False):
---> 93         super(ConvLSTM, self).__init__()
     94 
     95         self._check_kernel_size_consistency(kernel_size)

TypeError: super(type, obj): obj must be an instance or subtype of type

The other models (LSTM and CNN) can be loaded correctly multiple times.

Do you know what could be the issue?

Thanks

[iprapas]

This seems to be related to https://stackoverflow.com/questions/43751455/supertype-obj-obj-must-be-an-instance-or-subtype-of-type and https://github.com/jupyter/help/issues/130.

[RiccardoGrigoletto]

I solved the error with changing the line

https://github.com/DeepCube-org/uc3-public-notebooks/blob/45fea77035b8307afc6ab1602b30b1d698930271/src/modules/convlstm.py#L93

to super().__init__()

[RiccardoGrigoletto]

But you are correct, it's Jupyter that messes up with the module reloading.