def test_ermlp_predict(self): """Test ERMLP's predict function.""" ermlp = ERMLP(config=ERMLP_CONFIG) predictions = ermlp.predict(triples=TEST_TRIPLES) self.assertEqual(len(predictions), len(TEST_TRIPLES)) self.assertTrue(type(predictions.shape[0]), float)
def test_ermlp_predict(self): """Test ERMLP's predict function.""" ermlp = ERMLP(**ERMLP_CONFIG) ermlp.num_entities = ERMLP_CONFIG[NUM_ENTITIES] ermlp.num_relations = ERMLP_CONFIG[NUM_RELATIONS] predictions = ermlp.predict(triples=TEST_TRIPLES) self.assertEqual(len(predictions), len(TEST_TRIPLES)) self.assertTrue(type(predictions.shape[0]), float)
def test_instantiate_ermlp(self): """Test that ERMLP can be instantiated.""" ermlp = ERMLP(**ERMLP_CONFIG) ermlp.num_entities = ERMLP_CONFIG[NUM_ENTITIES] ermlp.num_relations = ERMLP_CONFIG[NUM_RELATIONS] self.assertIsNotNone(ermlp) self.assertEqual(ermlp.num_entities, 5) self.assertEqual(ermlp.num_relations, 5) self.assertEqual(ermlp.embedding_dim, 2) self.assertEqual(ermlp.margin_loss, 4)
def test_compute_scores_ermlp(self): """Test that SE's score function computes the scores correct.""" ermlp = ERMLP(config=ERMLP_CONFIG) h_embs = torch.tensor([[1., 1.], [1., 1.]], dtype=torch.float) r_embs = torch.tensor([[1., 1.], [2., 2.]], dtype=torch.float) t_embs = torch.tensor([[2., 2.], [4., 4.]], dtype=torch.float) scores = ermlp._compute_scores(h_embs, r_embs, t_embs).detach().cpu().numpy().tolist() self.assertEqual(len(scores), 2)
def test_instantiate_ermlp(self): """Test that ERMLP can be instantiated.""" ermlp = ERMLP(config=ERMLP_CONFIG) self.assertIsNotNone(ermlp) self.assertEqual(ermlp.num_entities, 5) self.assertEqual(ermlp.num_relations, 5) self.assertEqual(ermlp.embedding_dim, 2) self.assertEqual(ermlp.margin_loss, 4)