Using Locally Trained Sklearn Model in Algorithm

Hi Jonathan,

I read that you can persist your model via serialization, with pickle. See here : 

Now, you would need to serialize as a string. So from what I am reading, don't use joblib. Make sure your output is a string and save that to a file.

import pickle
s = pickle.dumps(my_model)
with open("my_model.txt", "w") as file:
    file.write(s)

Then you can most probably use QCAlgorithm's Download method and point it to your URI and get the string as output. The URI could I guess point to Dropbox, but make sure you add “?dl=1” at the end of the URI or in fact you will be downloading the web page showing your file to download, not the file itself.

And then use pickle to load.

my_model = self.Download("<Dropbox URI>?dl=1") # self is your QCAlgorithm class here - you need access to the algo object
clf = pickle.loads(my_model)
clf.predict(<...>)

Fred

P.S. Did not test. But if you did not, it's worth trying I'd say.