added test for link prediction

tests/test_general/test_link_prediction.py

@@ -1,13 +1,20 @@
-import pathlib
+import torch_geometric.transforms as T
+from stable_gnn.model_link_predict import ModelLinkPrediction
+from torch_geometric.datasets import Planetoid
+from sklearn.ensemble import GradientBoostingClassifier
+import pytest
 
-import torch
+@pytest.mark.parametrize("conv", ["SAGE", "GAT", "GCN"])
+@pytest.mark.parametrize("loss_name", ["APP", "LINE", "HOPE_AA", "VERSE_Adj"])
+def test_linkpredict(loss_name: str, conv: str) -> None:
+    root = '../tmp/'
+    name = 'Cora'
+    dataset = Planetoid(root=root + str(name), name=name, transform=T.NormalizeFeatures())
 
-from stable_gnn.graph import Graph
-from stable_gnn.pipelines.graph_classification_pipeline import TrainModelGC
-from tests.data_generators import generate_gc_graphs
+    model = ModelLinkPrediction(number_of_trials=50, loss_name=loss_name, emb_conv_name=conv)
 
-root = str(pathlib.Path(__file__).parent.resolve().joinpath("data_validation/")) + "/"
-generate_gc_graphs(root, 30)
+    train_edges, train_negative, test_edges, test_negative = model.train_test_edges(dataset)
 
-def test_linkpredict():
-    pass
+    cl_before = GradientBoostingClassifier(n_estimators=100, learning_rate=0.2, max_depth=5, random_state=0)
+    cl_after = model.train_cl(train_edges, train_negative)
+    assert (model.test(cl_before, test_edges, test_negative)) < (model.test(cl_after, test_edges, test_negative))