Loading deepchem/models/losses.py +2 −0 Original line number Diff line number Diff line Loading @@ -200,6 +200,8 @@ class SparseSoftmaxCrossEntropy(Loss): # This is for API consistency if len(output.shape) == 3: output = output.permute(0, 2, 1) if labels.shape[-1] == 1: labels = labels.squeeze(-1) return ce_loss(output, labels.long()) return loss Loading deepchem/models/tests/test_losses.py +6 −0 Original line number Diff line number Diff line Loading @@ -198,6 +198,12 @@ 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): """.""" Loading deepchem/models/torch_models/attentivefp.py +0 −5 Original line number Diff line number Diff line Loading @@ -306,9 +306,4 @@ class AttentiveFPModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(AttentiveFPModel, self)._prepare_batch( ([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights deepchem/models/torch_models/gat.py +0 −5 Original line number Diff line number Diff line Loading @@ -367,9 +367,4 @@ class GATModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(GATModel, self)._prepare_batch(([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights deepchem/models/torch_models/gcn.py +0 −5 Original line number Diff line number Diff line Loading @@ -351,9 +351,4 @@ class GCNModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(GCNModel, self)._prepare_batch(([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights Loading
deepchem/models/losses.py +2 −0 Original line number Diff line number Diff line Loading @@ -200,6 +200,8 @@ class SparseSoftmaxCrossEntropy(Loss): # This is for API consistency if len(output.shape) == 3: output = output.permute(0, 2, 1) if labels.shape[-1] == 1: labels = labels.squeeze(-1) return ce_loss(output, labels.long()) return loss Loading
deepchem/models/tests/test_losses.py +6 −0 Original line number Diff line number Diff line Loading @@ -198,6 +198,12 @@ 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): """.""" Loading
deepchem/models/torch_models/attentivefp.py +0 −5 Original line number Diff line number Diff line Loading @@ -306,9 +306,4 @@ class AttentiveFPModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(AttentiveFPModel, self)._prepare_batch( ([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights
deepchem/models/torch_models/gat.py +0 −5 Original line number Diff line number Diff line Loading @@ -367,9 +367,4 @@ class GATModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(GATModel, self)._prepare_batch(([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights
deepchem/models/torch_models/gcn.py +0 −5 Original line number Diff line number Diff line Loading @@ -351,9 +351,4 @@ class GCNModel(TorchModel): inputs = dgl.batch(dgl_graphs).to(self.device) _, labels, weights = super(GCNModel, self)._prepare_batch(([], labels, weights)) # torch.nn.CrossEntropy expects the last dimension of labels to be non-singleton if labels[0].shape[-1] == 1 and self.model.mode == 'classification': labels = [lbl.squeeze(-1) for lbl in labels] return inputs, labels, weights
