export.pkl file missing essential classes when uploaded to pythonanywhere : Forums : PythonAnywhere

Hello, I am trying to utilize a trained machine learning model in my website, which comes in the form of a .pkl file located in a directory named 'model'.

In a bash console, I am able to run: learn = load_learner('/home/AndrewGoldmann/mysite/static/model') This works.

The problem is when I run the same line in my website's app.py. I get the following error:

2020-07-07 14:03:45,204: An error occurred during a request.
Traceback (most recent call last):
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/app.py", line 2447, in wsgi_app
    response = self.full_dispatch_request()
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/app.py", line 1952, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/app.py", line 1821, in handle_user_exception
    reraise(exc_type, exc_value, tb)
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/_compat.py", line 39, in reraise
    raise value
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/app.py", line 1950, in full_dispatch_request
    rv = self.dispatch_request()
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/flask/app.py", line 1936, in dispatch_request
    return self.view_functions[rule.endpoint](**req.view_args)
  File "/home/AndrewGoldmann/mysite/app.py", line 182, in wrap
    return f(*args, **kwargs)
  File "/home/AndrewGoldmann/mysite/app.py", line 342, in api_receiveMarkup
    c.learn = load_learner('/home/AndrewGoldmann/mysite/static/model')
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/fastai/basic_train.py", line 622, in load_learner
    state = torch.load(source, map_location='cpu') if defaults.device == torch.device('cpu') else torch.load(source)
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/torch/serialization.py", line 593, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/home/AndrewGoldmann/.virtualenvs/flaskk/lib/python3.8/site-packages/torch/serialization.py", line 773, in _legacy_load
    result = unpickler.load()
AttributeError: Can't get attribute 'LabelSmoothingCrossEntropy' on <module '__main__' (built-in)>

Is there any explanation for why code run in bash console might produce a different result from what I run in app.py given that I am using the same virtual environment for both?

[edit by admin: formatting]

deleted-user-7739130 | 1 post | July 7, 2020, 2:13 p.m. | permalink