From 66e50aaa611f6d070bda056a954b90038ba5b5ff Mon Sep 17 00:00:00 2001
From: Darryl Ong <ongrongsheng@gmail.com>
Date: Wed, 18 Oct 2023 16:59:38 +0800
Subject: [PATCH] Added all items into test_dec_graph for all users

---
 cornac/models/gcmc/gcmc.py | 46 ++++++++++++++++++++++++++++++++++++--
 1 file changed, 44 insertions(+), 2 deletions(-)

diff --git a/cornac/models/gcmc/gcmc.py b/cornac/models/gcmc/gcmc.py
index 6b1229218..7bcaef0df 100644
--- a/cornac/models/gcmc/gcmc.py
+++ b/cornac/models/gcmc/gcmc.py
@@ -133,6 +133,44 @@ def _generate_dec_graph(data_set):
     )
 
 
+def _generate_test_dec_graph(data_set):
+    """
+    Generates decoding graph given a cornac data set
+
+    Parameters
+    ----------
+    data_set : cornac.data.dataset.Dataset
+        The data set as provided by cornac
+
+    Returns
+    -------
+    graph : dgl.heterograph
+        Heterograph containing user-item edges and nodes
+    """
+    uid_list = data_set.uir_tuple[0]
+    uid_list = np.unique(uid_list)
+
+    u_list = np.array([user_idx for _ in range(data_set.total_items) for user_idx in uid_list])
+    i_list = np.array([item_idx for item_idx in range(data_set.total_items) for _ in uid_list])
+
+    rating_pairs = (u_list, i_list)
+    ones = np.ones_like(rating_pairs[0])
+    user_item_ratings_coo = sp.coo_matrix(
+        (ones, rating_pairs),
+        shape=(data_set.total_users, data_set.total_items),
+        dtype=np.float32,
+    )
+
+    graph = dgl.bipartite_from_scipy(
+        user_item_ratings_coo, utype="_U", etype="_E", vtype="_V"
+    )
+
+    return dgl.heterograph(
+        {("user", "rate", "item"): graph.edges()},
+        num_nodes_dict={"user": data_set.total_users, "item": data_set.total_items},
+    )
+
+
 class Model:
     def __init__(
         self,
@@ -458,7 +496,7 @@ def predict(self, test_set):
             Dictionary containing '{user_idx}-{item_idx}' as key
             and {score} as value.
         """
-        test_dec_graph = _generate_dec_graph(test_set)
+        test_dec_graph = _generate_test_dec_graph(test_set)
         test_dec_graph = test_dec_graph.int().to(self.device)
 
         self.net.eval()
@@ -479,7 +517,11 @@ def predict(self, test_set):
 
         test_pred_ratings = test_pred_ratings.cpu().numpy()
 
-        (u_list, i_list, _) = test_set.uir_tuple
+        uid_list = test_set.uir_tuple[0]
+        uid_list = np.unique(uid_list)
+
+        u_list = np.array([user_idx for _ in range(test_set.total_items) for user_idx in uid_list])
+        i_list = np.array([item_idx for item_idx in range(test_set.total_items) for _ in uid_list])
 
         u_list = u_list.tolist()
         i_list = i_list.tolist()