Bug fix: in sine forward function, sum last dim.

piajiguai · Jul 11, 2022 · 6d2a45e · 6d2a45e
1 parent 03a9eba
commit 6d2a45e
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 9 deletions.
diff --git a/examples/matching/run_ml_sine.py b/examples/matching/run_ml_sine.py
@@ -48,9 +48,9 @@ def get_movielens_data(data_path, load_cache=False, seq_max_len=50):
                                                        mode=2,
                                                        neg_ratio=3,
                                                        min_item=0)
-        x_train = gen_model_input(df_train, user_profile, user_col, item_profile, item_col, seq_max_len=seq_max_len, padding='post', truncating='post')
+        x_train = gen_model_input(df_train, user_profile, user_col, item_profile, item_col, seq_max_len=seq_max_len, padding='pre', truncating='pre')
         y_train = np.array([0] * df_train.shape[0])  #label=0 means the first pred value is positive sample
-        x_test = gen_model_input(df_test, user_profile, user_col, item_profile, item_col, seq_max_len=seq_max_len, padding='post', truncating='post')
+        x_test = gen_model_input(df_test, user_profile, user_col, item_profile, item_col, seq_max_len=seq_max_len, padding='pre', truncating='pre')
         np.save("./data/ml-1m/saved/data_cache.npy", np.array((x_train, y_train, x_test), dtype=object))
 
     user_features, item_features, history_features, neg_item_features = ["user_id"], ["movie_id"], ["hist_movie_id"], ["neg_items"]
@@ -98,20 +98,20 @@ def main(dataset_path, model_name, epoch, learning_rate, batch_size, weight_deca
     parser = argparse.ArgumentParser()
     parser.add_argument('--dataset_path', default="./data/ml-1m/ml-1m_sample.csv")
     parser.add_argument('--model_name', default='sine')
-    parser.add_argument('--epoch', type=int, default=3)
-    parser.add_argument('--learning_rate', type=float, default=1e-4)
+    parser.add_argument('--epoch', type=int, default=10)
+    parser.add_argument('--learning_rate', type=float, default=1e-3)
     parser.add_argument('--batch_size', type=int, default=256)
-    parser.add_argument('--weight_decay', type=float, default=1e-4)
+    parser.add_argument('--weight_decay', type=float, default=1e-6)
     parser.add_argument('--device', default='cuda:0')
     parser.add_argument('--save_dir', default='./data/ml-1m/saved/')
     parser.add_argument('--seed', type=int, default=2022)    
 
     parser.add_argument('--embedding_dim', type=int, default=128)
     parser.add_argument('--hidden_dim', type=int, default=512)
-    parser.add_argument('--num_concept', type=int, default=50)
-    parser.add_argument('--num_intention', type=int, default=4)
+    parser.add_argument('--num_concept', type=int, default=10)
+    parser.add_argument('--num_intention', type=int, default=2)
     parser.add_argument('--temperature', type=int, default=0.1)
-    parser.add_argument('--seq_max_len', type=int, default=20)
+    parser.add_argument('--seq_max_len', type=int, default=50)
 
     args = parser.parse_args()
     main(args.dataset_path, args.model_name, args.epoch, args.learning_rate, args.batch_size, args.weight_decay, args.device,

diff --git a/torch_rechub/models/matching/sine.py b/torch_rechub/models/matching/sine.py
@@ -67,7 +67,7 @@ def forward(self, x):
         if self.mode == "item":
             return item_embedding
 
-        y = torch.mul(user_embedding, item_embedding).sum(dim=1)
+        y = torch.mul(user_embedding, item_embedding).sum(dim=-1)
 
         # # compute covariance regularizer
         # M = torch.cov(self.concept_embedding.weight, correction=0)