Loading a fine tuned model from disk fails

🤗Transformers
1

Hello!
I created my own model (i.e., class BertForMultiLabelClassification(BertPreTrainedModel)) and I train it with a transformers.Trainer. As training arguments I specify save_total_limit=2, load_best_model_at_end=True and save_strategy=epoch. After the training I have trainer.save_model(). This seems to work fine and I see the saved model and tokenizer files in the proper directory.
Loading and using the tokenizer with AutoTokenizer.from_pretrained(my_path) works fine.
Loading the model with AutoModel.from_pretrained(my_path) or BertForMultiLabelClassification.from_pretrained(my_path) fails with

ValueError: invalid literal for int() with base 10: 'CD000009'

May I ask for some help?
thank you very much!

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-62-3d86b810dabb> in <module>
----> 1 my_model = BertForMultiLabelClassification.from_pretrained(p)
      2 my_model

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/modeling_utils.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
   1198                 _from_auto=from_auto_class,
   1199                 _from_pipeline=from_pipeline,
-> 1200                 **kwargs,
   1201             )
   1202         else:

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/configuration_utils.py in from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
    460             )
    461 
--> 462         return cls.from_dict(config_dict, **kwargs)
    463 
    464     @classmethod

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/configuration_utils.py in from_dict(cls, config_dict, **kwargs)
    565         return_unused_kwargs = kwargs.pop("return_unused_kwargs", False)
    566 
--> 567         config = cls(**config_dict)
    568 
    569         if hasattr(config, "pruned_heads"):

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/models/bert/configuration_bert.py in __init__(self, vocab_size, hidden_size, num_hidden_layers, num_attention_heads, intermediate_size, hidden_act, hidden_dropout_prob, attention_probs_dropout_prob, max_position_embeddings, type_vocab_size, initializer_range, layer_norm_eps, pad_token_id, gradient_checkpointing, position_embedding_type, use_cache, **kwargs)
    141         **kwargs
    142     ):
--> 143         super().__init__(pad_token_id=pad_token_id, **kwargs)
    144 
    145         self.vocab_size = vocab_size

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/configuration_utils.py in __init__(self, **kwargs)
    266         if self.id2label is not None:
    267             kwargs.pop("num_labels", None)
--> 268             self.id2label = dict((int(key), value) for key, value in self.id2label.items())
    269             # Keys are always strings in JSON so convert ids to int here.
    270         else:

~/.azureml/envs/azureml_389e19432b898fec54de7429d4696827/lib/python3.6/site-packages/transformers/configuration_utils.py in <genexpr>(.0)
    266         if self.id2label is not None:
    267             kwargs.pop("num_labels", None)
--> 268             self.id2label = dict((int(key), value) for key, value in self.id2label.items())
    269             # Keys are always strings in JSON so convert ids to int here.
    270         else:

ValueError: invalid literal for int() with base 10: 'CD000009'