#53 partially done. Need to add load model

API/app.py

@@ -35,9 +35,9 @@ def get_recommended_movies(user_id):
     also triggers sending recommender outputs to front
     """
     from recommend_movies import RecommendMovies
-    return RecommendMovies(user_id, None, None).recommend()
+    return RecommendMovies(user_id).recommend()
 
 @app.post("/rating")
 def post_user_rating(user_id, movie_id, rating):
     from recommend_movies import RecommendMovies
-    return RecommendMovies(user_id, None, None).rate_movie(movie_id, rating)
+    return RecommendMovies(user_id).rate_movie(movie_id, rating)

API/connect_db.py

@@ -55,3 +55,19 @@ def delete_user_entry(self, user_id, movie_id):
 
     def get_movie(self, movie_id):
         return self.get_one_info({"movie_id": movie_id})
+
+    def get_sequential_model(self):
+        return self.get_info({"model":"sequential"})
+
+    def get_explicit_model(self):
+        return self.get_info({"model": "explicit"})
+
+    def load_db_access_credentials(self):
+        with open("./credentials.txt", "r") as fh:
+            for line in fh:
+                if "user=" in line:
+                    user = line.strip().split("=")[-1]
+                    self.db_access_key = user
+                elif "pass=" in line:
+                    passw = line.strip().split("=")[-1]
+                    self.db_pass = passw

API/recommend_movies.py

@@ -1,22 +1,25 @@
+import os.path
+
+import torch
 from connect_db import MongoDB
 from typing import List, Dict
 class RecommendMovies:
-    MONGO_ACCESS = MongoDB()
+    MONGO_ACCESS = MongoDB(collection="models")
 
-    def __init__(self, user_id, explicit_model, sequential_model):
+    def __init__(self, user_id):
         self.user_id = user_id
         self.user_data:List[Dict] = self.get_user_data(user_id)
-        self.explicit_model = explicit_model
-        self.sequential_model= sequential_model
+        self.explicit_model = self.load_explicit_model()
+        self.sequential_model= self.load_sequential_model()
 
     def recommend(self):
         if not self.user_data:
             return None
         previous_movies = self.get_previously_watched_movies()
         exp_recommend_movies = self.get_recommendations_from_explicit_model()
         seq_recommend_movies = self.get_recommendations_from_sequential_model()
-        #exp_recommend_movies = set()
-        filtered_movies = self.filter_previously_watched_movies(exp_recommend_movies, previous_movies)
+        recommended_movies = exp_recommend_movies + seq_recommend_movies
+        filtered_movies = self.filter_previously_watched_movies(recommended_movies, previous_movies)
         return self.get_movie_title_and_genre(filtered_movies)
 
     def rate_movie(self, movie_id, rating):
@@ -27,17 +30,17 @@ def rate_movie(self, movie_id, rating):
     def get_user_data(self,user_id):
         return self.MONGO_ACCESS.get_user(user_id)
 
-    def get_recommendations_from_explicit_model(self):
-        ##TODO: load explicit model, ask user_id, get movie ids
-        ##TODO: connect it to the explicit model thing
-        return ["12", "11", "99"]
-        #return self.explicit_model(self.user_id)
-
+    def get_recommendations_from_explicit_model(self, n_movies=10):
+        import numpy as np
+        from process_data.process_movie_data import MovieLens
+        dataset=MovieLens.download_data_from_db()
+        ratings = self.sequential_model.predict(user_ids=self.user_id)
+        indices = np.argpartition(ratings, -n_movies)[-n_movies:]
+        best_movie_ids = indices[np.argsort(ratings[indices])]
+        return [dataset.item_ids[i] for i in best_movie_ids]
 
     def get_recommendations_from_sequential_model(self):
-        ##TODO: load sequential model, ask user_id, get movie ids
-        pass
-        #return self.sequential_model(self.user_id)
+        return self.sequential_model(self.user_id)
 
     def get_previously_watched_movies(self):
         return [movie_data["movie_id"] for movie_data in self.user_data]
@@ -46,6 +49,16 @@ def get_previously_watched_movies(self):
     def filter_previously_watched_movies(recommended_movies, previous_movies):
         return list(set(recommended_movies).difference(set(previous_movies)))
 
+    def load_sequential_model(self):
+        seq_model_path = self.MONGO_ACCESS.get_sequential_model()
+        return torch.load(seq_model_path)
+
+    def load_explicit_model(self):
+        #exp_model_path = self.MONGO_ACCESS.get_explicit_model()
+        from config import DATA_PATH
+        exp_model_path = os.path.join(DATA_PATH, "explicit.pt")
+        return torch.load(exp_model_path)
+
     def get_movie_title_and_genre(self, movie_ids:List):
         recommended_movies = []
         for movie_id in movie_ids:

requirements.txt

@@ -7,6 +7,7 @@ uvicorn
 fastapi
 pymongo[srv]
 pytest
+mlflow
 h5py
 tables
 cython