change type to redis

personas_vector/to_vector.py

@@ -53,12 +53,12 @@ if __name__ == "__main__":
     tags_vector_dict = {}
     for i in tags_set:
         try:
-            # vec = json.dumps(model.wv.get_vector(i).tolist())
-            vec = model.wv.get_vector(i)
+            vec = json.dumps(model.wv.get_vector(i).tolist())
+            # vec = model.wv.get_vector(i)
             tags_vector_dict[i] = vec
         except Exception as e:
             pass
-    redis_client_db.hmset("personas_tags_embedding", tags_vector_dict)
+    # redis_client_db.hmset("personas_tags_embedding", tags_vector_dict)
 
     print(len(tags_vector_dict.items()))
     # print(random.choice(list(tags_vector_dict.items())))
@@ -68,9 +68,9 @@ if __name__ == "__main__":
     for _, row in tractate_tags_df.iterrows():
         vecs = []
         for i in row["business_tags"]:
-            vec = tags_vector_dict.get(i, np.array([]))
-            if vec.any():
-                vecs.append(vec)
+            vec = tags_vector_dict.get(i)
+            if vec:
+                vecs.append(np.array(json.loads(vec)).astype("float32"))
         if vecs:
             tractate_vector_dict[row["tractate_id"]] = np.average(vecs, axis=0)
     print(len(tractate_vector_dict.items()))
@@ -91,20 +91,21 @@ if __name__ == "__main__":
     base_dir = os.getcwd()
     model_dir = os.path.join(base_dir, "_models")
     index_path = os.path.join(model_dir, "faiss_personas_vector.index")
-    faiss.write_index(index2, index_path)
+    # faiss.write_index(index2, index_path)
     print(index_path)
 
-    # device vector
-    # for _, row in device_tags_df.iterrows():
-    #     vecs = []
-    #     for i in row["business_tags"]:
-    #         vec = tags_vector_dict.get(i, np.array([]))
-    #         if vec.any():
-    #             vecs.append(vec)
-    #     if vecs:
-    #         t = np.array([np.average(vecs, axis=0)]).astype("float32")
-    #         D, I = index2.search(t, 10)
-    #         print(row["cl_id"], row["business_tags"])
-    #         print(I)
+    device vector
+    for _, row in device_tags_df.iterrows():
+        vecs = []
+        for i in row["business_tags"]:
+            # vec = tags_vector_dict.get(i, np.array([]))
+            vec = tags_vector_dict.get(i)
+            if vec:
+                vecs.append(np.array(json.loads(vec)).astype("float32"))
+        if vecs:
+            t = np.array([np.average(vecs, axis=0)]).astype("float32")
+            D, I = index2.search(t, 10)
+            print(row["cl_id"], row["business_tags"])
+            print(I)
 
 # curl "http://172.16.31.17:9000/gm-dbmw-tractate-read/_search?pretty" -d '{"query": {"term": {"id": "10269"}}, "_source": {"include": ["content", "portrait_tag_name"]}}'