在子包中加载一个pickle文件->模块'__main__'错误

Question

I get this error message when I try to load a pickle file:

Can't get attribute 'CustomNMF' on <module '__main__' from '/Users/cricket/Documents/ONline classes/A6 - Stackoverflow/Tag-generator-Stackoverflow/app.py'

My question where am I missing an import CustomNMF or any other thing that would make it work ?

It's a flask project I'm working on, the structure is:

project folders:

+ app.py
+ app (sub-folder)
++ __init__.py
++ views.py
++ CustomNMF.py
++ and other things...

Content of app.py :

from app import app
if __name__ == "__main__":
    app.run(debug=True)

content of __init__.py

from flask import Flask
app = Flask(__name__)
from app import views

Content of views.py :

....
from app.CustomNMF import CustomNMF
customNMF = CustomNMF()  #That works fine !

def loadpickle(filename = 'customNMF.pickle'):
    with open(os.path.join(os.getcwd(),'app', filename), 'rb') as f:
        return pickle.load(f)

customNMF = loadpickle('customNMF.pickle')  #Line that is generating the error :-(
....

Content of CustomNMF.py : (A simple copy paste from the notebook used to create the class )

class CustomNMF(BaseEstimator, TransformerMixin):
    def __init__(self,popularitytags=None, **params): 
        self.popularitytags = popularitytags
        .......

I'm sure it's obvious for most of you, and the solution is just a small detail I'm missing. Any help will be appreciated.

Answer 1

I think I found the issue: I saved the customNMF trained from the python notebook. I had to slightly modify the classes in the dask version not not call some unnecessary packages (cross validation and metrics from sklearn ) for the website version. I assume those minor modifications are the source of the problem.

Now the CustomNMF is trained and pickled for later usage by the website server. Not sure it's the best approach, but it's working!

If anyone as good lecture about packages structure and how to articulate them (simplified version ;-) just for me), I would really appreciate.

Thanks