fix bug

benchmark/tgt_ijcai_benchmark.py

@@ -401,7 +401,7 @@ def test(model: torch.nn.Module, loader: NeighborLoader, desc: str, stage: str,
         input_time=torch.full((num_src_nodes, ), split_date[split],
                               dtype=torch.long),
         subgraph_type="bidirectional",
-        batch_size=args.batch_size,
+        batch_size=train_search_space['batch_size'],
         temporal_strategy=args.temporal_strategy,
         shuffle=split == "train",
         num_workers=args.num_workers,
@@ -460,7 +460,8 @@ def train_and_eval_with_cfg(
         print(f"Train Loss: {train_loss:.4f}")
         if epoch % 5 == 0:
             # Check if we should early stop
-            test_pred = test(model, loader_dict["test"], "test", "test")
+            test_pred = test(model, loader_dict["test"], "test", "test",
+                             target=target_list)
             test_metric = calculate_hit_rate(test_pred, target_list)
             print(f"Test metric: {test_metric:.4f}")
             if test_metric > best_test_metric: