Merge pull request #2328 from mufeili/master (50992082) · Commits · 钟慕尧 / deepchem

deepchem/models/losses.py

+8 −0

Original line number	Diff line number	Diff line
		@@ -187,7 +187,12 @@ class SparseSoftmaxCrossEntropy(Loss):

		def _compute_tf_loss(self, output, labels):
		import tensorflow as tf

		if len(labels.shape) == len(output.shape):
		labels = tf.squeeze(labels, axis=-1)

		labels = tf.cast(labels, tf.int32)

		return tf.nn.sparse_softmax_cross_entropy_with_logits(labels, output)

		def _create_pytorch_loss(self):
		@@ -200,6 +205,9 @@ class SparseSoftmaxCrossEntropy(Loss):
		# This is for API consistency
		if len(output.shape) == 3:
		output = output.permute(0, 2, 1)

		if len(labels.shape) == len(output.shape):
		labels = labels.squeeze(-1)
		return ce_loss(output, labels.long())

		return loss

+27 −0

Original line number	Diff line number	Diff line
		@@ -34,6 +34,19 @@ def test_attentivefp_regression():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean_absolute_error'] < 0.5

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_delaney

		tasks, all_dataset, transformers = load_delaney(featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = AttentiveFPModel(
		mode='regression',
		n_tasks=len(tasks),
		num_layers=1,
		num_timesteps=1,
		graph_feat_size=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')
		@@ -56,6 +69,20 @@ def test_attentivefp_classification():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean-roc_auc_score'] >= 0.85

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_bace_classification

		tasks, all_dataset, transformers = load_bace_classification(
		featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = AttentiveFPModel(
		mode='classification',
		n_tasks=len(tasks),
		num_layers=1,
		num_timesteps=1,
		graph_feat_size=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')

+29 −0

Original line number	Diff line number	Diff line
		@@ -39,6 +39,20 @@ def test_gat_regression():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean_absolute_error'] < 0.5

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_delaney

		tasks, all_dataset, transformers = load_delaney(featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = dc.models.GATModel(
		mode='regression',
		n_tasks=len(tasks),
		graph_attention_layers=[2],
		n_attention_heads=1,
		residual=False,
		predictor_hidden_feats=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')
		@@ -62,6 +76,21 @@ def test_gat_classification():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean-roc_auc_score'] >= 0.85

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_bace_classification

		tasks, all_dataset, transformers = load_bace_classification(
		featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = dc.models.GATModel(
		mode='classification',
		n_tasks=len(tasks),
		graph_attention_layers=[2],
		n_attention_heads=1,
		residual=False,
		predictor_hidden_feats=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')

+26 −0

Original line number	Diff line number	Diff line
		@@ -39,6 +39,18 @@ def test_gcn_regression():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean_absolute_error'] < 0.5

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_delaney

		tasks, all_dataset, transformers = load_delaney(featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = dc.models.GCNModel(
		n_tasks=len(tasks),
		graph_conv_layers=[2],
		residual=False,
		predictor_hidden_feats=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')
		@@ -62,6 +74,20 @@ def test_gcn_classification():
		scores = model.evaluate(dataset, [metric], transformers)
		assert scores['mean-roc_auc_score'] >= 0.85

		# test on a small MoleculeNet dataset
		from deepchem.molnet import load_bace_classification

		tasks, all_dataset, transformers = load_bace_classification(
		featurizer=featurizer)
		train_set, _, _ = all_dataset
		model = dc.models.GCNModel(
		mode='classification',
		n_tasks=len(tasks),
		graph_conv_layers=[2],
		residual=False,
		predictor_hidden_feats=2)
		model.fit(train_set, nb_epoch=1)


		@unittest.skipIf(not has_torch_and_dgl,
		'PyTorch, DGL, or DGL-LifeSci are not installed')

+12 −0

Original line number	Diff line number	Diff line
		@@ -186,6 +186,12 @@ class TestLosses(unittest.TestCase):
		expected = [-np.log(softmax[0, 1]), -np.log(softmax[1, 0])]
		assert np.allclose(expected, result)

		labels = tf.constant([[1], [0]])
		result = loss._compute_tf_loss(outputs, labels).numpy()
		softmax = np.exp(y) / np.expand_dims(np.sum(np.exp(y), axis=1), 1)
		expected = [-np.log(softmax[0, 1]), -np.log(softmax[1, 0])]
		assert np.allclose(expected, result)

		@unittest.skipIf(not has_pytorch, 'PyTorch is not installed')
		def test_sparse_softmax_cross_entropy_pytorch(self):
		"""Test SparseSoftmaxCrossEntropy."""
		@@ -198,6 +204,12 @@ class TestLosses(unittest.TestCase):
		expected = [-np.log(softmax[0, 1]), -np.log(softmax[1, 0])]
		assert np.allclose(expected, result)

		labels = torch.tensor([[1], [0]])
		result = loss._create_pytorch_loss()(outputs, labels).numpy()
		softmax = np.exp(y) / np.expand_dims(np.sum(np.exp(y), axis=1), 1)
		expected = [-np.log(softmax[0, 1]), -np.log(softmax[1, 0])]
		assert np.allclose(expected, result)

		@unittest.skipIf(not has_tensorflow, 'TensorFlow is not installed')
		def test_VAE_ELBO_tf(self):
		"""."""