Reputation: 1934
I'm trying to persist prediction model state by saving my model with
joblib.dump(model, path)
This works fine, but if I do it twice, the second set of generated .pkl_* files don't really overwrite the old ones, so when I go to call joblib.load(path), I can't be sure it's actually loading in the correct model.
How do I make it so joblib.dump destroys the old .pkl_* files before saving new ones?
Upvotes: 7
Views: 2490
Reputation: 119
You actually don't need to do the "wb" while using joblib. I was facing the similar issue; realising later that I was dumping and loading from different locations.
Upvotes: 0
Reputation: 2831
Instead of a path if you pass in a file opened with "wb" then it will overwrite.
Upvotes: 2