cistrome/MIRA

Errors when loading trained models

Closed this issue · 4 comments

KongMingxi commented

Hi authors,

Many thanks for your excellent work and contributions! I am learning the tutorial "2021-12-26_MIRA_tutorial.ipynb".
When I load the trained model using mira.topic_model.ExpressionTopicModel.load(), I got this error:

UnpicklingError Traceback (most recent call last)
/tmp/ipykernel_47536/3391690466.py in
----> 1 tt.load('./models/best_rna_model.pth')

~/miniconda3/envs/mira/lib/python3.7/site-packages/mira/topic_model/base.py in load(cls, filename)
165 '''
166
--> 167 data = torch.load(filename)
168
169 model = cls(**data['params'])

~/miniconda3/envs/mira/lib/python3.7/site-packages/torch/serialization.py in load(f, map_location, pickle_module, **pickle_load_args)
606 return torch.jit.load(opened_file)
607 return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
--> 608 return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
609
610

~/miniconda3/envs/mira/lib/python3.7/site-packages/torch/serialization.py in _legacy_load(f, map_location, pickle_module, **pickle_load_args)
775 "functionality.")
776
--> 777 magic_number = pickle_module.load(f, **pickle_load_args)
778 if magic_number != MAGIC_NUMBER:
779 raise RuntimeError("Invalid magic number; corrupt file?")

UnpicklingError: NEWOBJ expected an arg tuple.

I also tried to load it using torch.load() but it doesn't work.

I am wondering how I can load trained models (.pth files)?

Best,
Mingxi

AllenWLynch commented
KongMingxi commented

Hi AL,

Thanks for your help! Yes, I specified the path for the .pth file in my "load" command.
My code for model building, saving and loading is as follows:

  1. to initialize a model for scRNA data:
    example_rna_model = mira.topics.ExpressionTopicModel(exogenous_key='exog_feature', endogenous_key='highly_variable')
  2. to train and save the best model after hyperparameter tuning:
    tuner = mira.topics.TopicModelTuner(example_rna_model, save_name="./my_folder/best_rna_model.pth")
    tuner.train_test_split(rna_data, train_size = 0.8)
    study = tuner.tune(rna_data)
    example_rna_model = tuner.select_best_model(rna_data)
  3. to load the trained model:
    rna_model_trained = mira.topic_model.ExpressionTopicModel.load('./my_folder/best_rna_model.pth')

Also, I tried to load the trained model using the following command, but it failed to load:
rna_model_trained = torch.load('./my_folder/best_rna_model.pth')

Please let me know if you need more code.

Thanks!
Mingxi

AllenWLynch commented
KongMingxi commented

Thanks! I can save and load models now!
Thanks again for your help!