refinement and debug (2dd25e25) · Commits · 钟慕尧 / deepchem

deepchem/models/tensorgraph/graph_layers.py

+38 −36

Original line number	Diff line number	Diff line
		@@ -75,15 +75,15 @@ class WeaveLayer(Layer):
		"""
		Parameters
		----------
		n_atom_input_feat: int
		n_atom_input_feat: int, optional
		Number of features for each atom in input.
		n_pair_input_feat: int
		n_pair_input_feat: int, optional
		Number of features for each pair of atoms in input.
		n_atom_output_feat: int
		n_atom_output_feat: int, optional
		Number of features for each atom in output.
		n_pair_output_feat: int
		n_pair_output_feat: int, optional
		Number of features for each pair of atoms in output.
		n_hidden_XX: int
		n_hidden_XX: int, optional
		Number of units(convolution depths) in corresponding hidden layer
		update_pair: bool, optional
		Whether to calculate for pair features,
		@@ -96,6 +96,7 @@ class WeaveLayer(Layer):
		Dropout probability, not supported here

		"""
		super(WeaveLayer, self).__init__(**kwargs)
		self.init = initializations.get(init) # Set weight initialization
		self.activation = activations.get(activation) # Get activations
		self.update_pair = update_pair # last weave layer does not need to update
		@@ -111,7 +112,6 @@ class WeaveLayer(Layer):
		self.n_atom_output_feat = n_atom_output_feat
		self.n_pair_output_feat = n_pair_output_feat
		self.W_AP, self.b_AP, self.W_PP, self.b_PP, self.W_P, self.b_P = None, None, None, None, None, None
		super(WeaveLayer, self).__init__(**kwargs)

		def build(self):
		""" Construct internal trainable weights.
		@@ -200,6 +200,7 @@ class WeaveLayer(Layer):
		P = self.activation(P)
		else:
		P = pair_features

		out_tensor = [A, P]
		if set_tensors:
		self.variables = self.trainable_weights
		@@ -295,6 +296,7 @@ class WeaveGather(Layer):
		if self.gaussian_expand:
		output_molecules = tf.matmul(output_molecules, self.W) + self.b
		output_molecules = self.activation(output_molecules)

		out_tensor = output_molecules
		if set_tensors:
		self.variables = self.trainable_weights

deepchem/models/tensorgraph/layers.py

+46 −2

Original line number	Diff line number	Diff line
		@@ -5,7 +5,7 @@ from collections import Sequence
		import tensorflow as tf
		import numpy as np

		from deepchem.nn import model_ops, initializations
		from deepchem.nn import model_ops, initializations, regularizers, activations


		class Layer(object):
		@@ -489,7 +489,6 @@ class L2Loss(Layer):
		return out_tensor



		class SoftMax(Layer):

		def __init__(self, **kwargs):
		@@ -1077,6 +1076,51 @@ class BatchNorm(Layer):
		self.out_tensor = out_tensor
		return out_tensor

		class BatchNormalization(Layer):
		def __init__(self,
		epsilon=1e-5,
		axis=-1,
		momentum=0.99,
		beta_init='zero',
		gamma_init='one',
		**kwargs):
		self.beta_init = initializations.get(beta_init)
		self.gamma_init = initializations.get(gamma_init)
		self.epsilon = epsilon
		self.axis = axis
		self.momentum = momentum
		super(BatchNormalization, self).__init__(**kwargs)

		def add_weight(self, shape, initializer, name=None):
		initializer = initializations.get(initializer)
		weight = initializer(shape, name=name)
		return weight

		def build(self, input_shape):
		shape = (input_shape[self.axis],)
		self.gamma = self.add_weight(
		shape,
		initializer=self.gamma_init,
		name='{}_gamma'.format(self.name))
		self.beta = self.add_weight(
		shape,
		initializer=self.beta_init,
		name='{}_beta'.format(self.name))

		def create_tensor(self, in_layers=None, set_tensors=True, **kwargs):
		inputs = self._get_input_tensors(in_layers)
		x = inputs[0]
		input_shape = model_ops.int_shape(x)
		self.build(input_shape)
		m = model_ops.mean(x, axis=-1, keepdims=True)
		std = model_ops.sqrt(model_ops.var(x, axis=-1, keepdims=True) + self.epsilon)
		x_normed = (x - m) / (std + self.epsilon)
		x_normed = self.gamma * x_normed + self.beta
		out_tensor = x_normed
		if set_tensors:
		self.out_tensor = out_tensor
		return out_tensor


		class WeightedError(Layer):

deepchem/models/tensorgraph/models/graph_models.py

+5 −6

Original line number	Diff line number	Diff line
		@@ -7,7 +7,7 @@ from deepchem.metrics import to_one_hot, from_one_hot
		from deepchem.models.tensorgraph.graph_layers import WeaveLayer, WeaveGather, \
		Combine_AP, Separate_AP, DTNNEmbedding, DTNNStep, DTNNGather, DAGLayer, DAGGather, DTNNExtract
		from deepchem.models.tensorgraph.layers import Dense, Concat, SoftMax, SoftMaxCrossEntropy, GraphConv, BatchNorm, \
		GraphPool, GraphGather, WeightedError
		GraphPool, GraphGather, WeightedError, BatchNormalization
		from deepchem.models.tensorgraph.layers import L2Loss, Label, Weights, Feature
		from deepchem.models.tensorgraph.tensor_graph import TensorGraph
		from deepchem.trans import undo_transforms
		@@ -72,9 +72,9 @@ class WeaveTensorGraph(TensorGraph):
		separated = Separate_AP(in_layers=[weave_layer2])
		dense1 = Dense(
		out_channels=self.n_graph_feat,
		activation_fn=tf.nn.relu,
		activation_fn=tf.nn.tanh,
		in_layers=[separated])
		batch_norm1 = BatchNorm(in_layers=[dense1])
		batch_norm1 = BatchNormalization(epsilon=1e-5, mode=1, in_layers=[dense1])
		weave_gather = WeaveGather(
		self.batch_size,
		n_input=self.n_graph_feat,
		@@ -103,7 +103,6 @@ class WeaveTensorGraph(TensorGraph):
		self.labels_fd.append(label)
		cost = L2Loss(in_layers=[label, regression])
		costs.append(cost)

		all_cost = Concat(in_layers=costs, axis=1)
		self.weights = Weights(shape=(None, self.n_tasks))
		loss = WeightedError(in_layers=[all_cost, self.weights])
		@@ -162,7 +161,7 @@ class WeaveTensorGraph(TensorGraph):

		feed_dict[self.atom_features] = np.concatenate(atom_feat, axis=0)
		feed_dict[self.pair_features] = np.concatenate(pair_feat, axis=0)
		feed_dict[self.pair_split] = np.array(pair_split)
		feed_dict[self.pair_split] = pair_split
		feed_dict[self.atom_split] = np.array(atom_split)
		feed_dict[self.atom_to_pair] = np.concatenate(atom_to_pair, axis=0)
		yield feed_dict
		@@ -398,7 +397,7 @@ class DAGTensorGraph(TensorGraph):
		cost = L2Loss(in_layers=[label, regression])
		costs.append(cost)

		all_cost = Concat(in_layers=costs)
		all_cost = Concat(in_layers=costs, axis=1)
		self.weights = Weights(shape=(None, self.n_tasks))
		loss = WeightedError(in_layers=[all_cost, self.weights])
		self.set_loss(loss)

deepchem/models/tensorgraph/tensor_graph.py

+2 −1

Original line number	Diff line number	Diff line
		@@ -69,7 +69,8 @@ class TensorGraph(Model):
		tf.train.AdamOptimizer,
		learning_rate=learning_rate,
		beta1=0.9,
		beta2=0.999)
		beta2=0.999,
		epsilon=1e-7)

		# Singular place to hold Tensor objects which don't serialize
		# These have to be reconstructed on restoring from pickle

examples/delaney/delaney_tensorgraph_DAG.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -44,7 +44,7 @@ model = dc.models.DAGTensorGraph(
		mode='regression')

		# Fit trained model
		model.fit(train_dataset, nb_epoch=50)
		model.fit(train_dataset, nb_epoch=50, checkpoint_interval=100)
		print("Evaluating model")
		train_scores = model.evaluate(train_dataset, [metric], transformers)
		valid_scores = model.evaluate(valid_dataset, [metric], transformers)

Admin message