Clear cached properties on partial_fit_* (#685)

The item_norms/user_norms were incorrect after calling the partial_fit methods, especially for new items. Fix.

Author: benfred

implicit/cpu/als.py

@@ -306,6 +306,10 @@ def partial_fit_users(self, userids, user_items):
         # update the stored factors with the newly calculated values
         self.user_factors[userids] = user_factors
 
+        # clear any cached properties that are invalidated by this update
+        self._user_norms = None
+        self._XtX = None
+
     def partial_fit_items(self, itemids, item_users):
         """Incrementally updates item factors
 
@@ -339,6 +343,10 @@ def partial_fit_items(self, itemids, item_users):
         # update the stored factors with the newly calculated values
         self.item_factors[itemids] = item_factors
 
+        # clear any cached properties that are invalidated by this update
+        self._item_norms = None
+        self._YtY = None
+
     def explain(self, userid, user_items, itemid, user_weights=None, N=10):
         """Provides explanations for why the item is liked by the user.
 

implicit/gpu/als.py

@@ -234,6 +234,10 @@ def partial_fit_users(self, userids, user_items):
 
         self.user_factors.assign_rows(userids, user_factors)
 
+        # clear any cached properties that are invalidated by this update
+        self._user_norms = self._user_norms_host = None
+        self._XtX = None
+
     def partial_fit_items(self, itemids, item_users):
         """Incrementally updates item factors
 
@@ -266,6 +270,10 @@ def partial_fit_items(self, itemids, item_users):
         # update the stored factors with the newly calculated values
         self.item_factors.assign_rows(itemids, item_factors)
 
+        # clear any cached properties that are invalidated by this update
+        self._item_norms = self._item_norms_host = None
+        self._YtY = None
+
     @property
     def solver(self):
         if self._solver is None: