Finished refactoring graph models (41dea7db) · Commits · 钟慕尧 / deepchem

deepchem/models/tensorgraph/graph_layers.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -286,7 +286,7 @@ class WeaveGather(Layer):
		dist_max = [dist[i].prob(gaussian_memberships[i][0]) for i in range(11)]
		outputs = [dist[i].prob(x) / dist_max[i] for i in range(11)]
		outputs = tf.stack(outputs, axis=2)
		outputs = outputs / tf.reduce_sum(outputs, axis=2, keep_dims=True)
		outputs = outputs / tf.reduce_sum(outputs, axis=2, keepdims=True)
		outputs = tf.reshape(outputs, [-1, self.n_input * 11])
		return outputs

deepchem/models/tensorgraph/models/graph_models.py

+23 −35

Original line number	Diff line number	Diff line
		@@ -156,8 +156,6 @@ class WeaveModel(TensorGraph):
		pad_batches=True):
		"""TensorGraph style implementation """
		for epoch in range(epochs):
		if not predict:
		print('Starting epoch %i' % epoch)
		for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(
		batch_size=self.batch_size,
		deterministic=deterministic,
		@@ -221,6 +219,7 @@ class DTNNModel(TensorGraph):
		distance_max=18,
		output_activation=True,
		mode="regression",
		dropout=0.0,
		**kwargs):
		"""
		Parameters
		@@ -240,6 +239,8 @@ class DTNNModel(TensorGraph):
		maximum distance of atom pairs, default = 18 Angstorm
		mode: str
		Either "classification" or "regression" for type of model.
		dropout: float
		the dropout probablity to use.
		"""
		if mode not in ['classification', 'regression']:
		raise ValueError("mode must be either 'classification' or 'regression'")
		@@ -255,6 +256,7 @@ class DTNNModel(TensorGraph):
		self.steps = np.expand_dims(self.steps, 0)
		self.output_activation = output_activation
		self.mode = mode
		self.dropout = dropout
		super(DTNNModel, self).__init__(**kwargs)
		assert self.mode == "regression"
		self.build_graph()
		@@ -271,6 +273,8 @@ class DTNNModel(TensorGraph):

		dtnn_embedding = DTNNEmbedding(
		n_embedding=self.n_embedding, in_layers=[self.atom_number])
		if self.dropout > 0.0:
		dtnn_embedding = Dropout(self.dropout, in_layers=dtnn_embedding)
		dtnn_layer1 = DTNNStep(
		n_embedding=self.n_embedding,
		n_distance=self.n_distance,
		@@ -278,6 +282,8 @@ class DTNNModel(TensorGraph):
		dtnn_embedding, self.distance, self.distance_membership_i,
		self.distance_membership_j
		])
		if self.dropout > 0.0:
		dtnn_layer1 = Dropout(self.dropout, in_layers=dtnn_layer1)
		dtnn_layer2 = DTNNStep(
		n_embedding=self.n_embedding,
		n_distance=self.n_distance,
		@@ -285,27 +291,26 @@ class DTNNModel(TensorGraph):
		dtnn_layer1, self.distance, self.distance_membership_i,
		self.distance_membership_j
		])
		if self.dropout > 0.0:
		dtnn_layer2 = Dropout(self.dropout, in_layers=dtnn_layer2)
		dtnn_gather = DTNNGather(
		n_embedding=self.n_embedding,
		layer_sizes=[self.n_hidden],
		n_outputs=self.n_tasks,
		output_activation=self.output_activation,
		in_layers=[dtnn_layer2, self.atom_membership])
		if self.dropout > 0.0:
		dtnn_gather = Dropout(self.dropout, in_layers=dtnn_gather)

		costs = []
		self.labels_fd = []
		for task in range(self.n_tasks):
		regression = DTNNExtract(task, in_layers=[dtnn_gather])
		self.add_output(regression)
		label = Label(shape=(None, 1))
		self.labels_fd.append(label)
		cost = L2Loss(in_layers=[label, regression])
		costs.append(cost)

		all_cost = Stack(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)
		n_tasks = self.n_tasks
		weights = Weights(shape=(None, n_tasks))
		labels = Label(shape=(None, n_tasks))
		output = Reshape(
		shape=(None, n_tasks),
		in_layers=[Dense(in_layers=dtnn_gather, out_channels=n_tasks)])
		self.add_output(output)
		weighted_loss = ReduceSum(L2Loss(in_layers=[labels, output, weights]))
		self.set_loss(weighted_loss)

		def default_generator(self,
		dataset,
		@@ -315,8 +320,6 @@ class DTNNModel(TensorGraph):
		pad_batches=True):
		"""TensorGraph style implementation"""
		for epoch in range(epochs):
		if not predict:
		print('Starting epoch %i' % epoch)
		for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(
		batch_size=self.batch_size,
		deterministic=deterministic,
		@@ -324,10 +327,9 @@ class DTNNModel(TensorGraph):

		feed_dict = dict()
		if y_b is not None:
		for index, label in enumerate(self.labels_fd):
		feed_dict[label] = y_b[:, index:index + 1]
		feed_dict[self.labels[0]] = y_b
		if w_b is not None:
		feed_dict[self.weights] = w_b
		feed_dict[self.task_weights[0]] = w_b
		distance = []
		atom_membership = []
		distance_membership_i = []
		@@ -363,18 +365,6 @@ class DTNNModel(TensorGraph):

		yield feed_dict

		def predict(self, dataset, transformers=[], outputs=None):
		if outputs is None:
		outputs = self.outputs
		if transformers != [] and not isinstance(outputs, collections.Sequence):
		raise ValueError(
		"DTNN does not support single tensor output with transformers")
		retval = super(DTNNModel, self).predict(dataset, outputs=outputs)
		if not isinstance(outputs, collections.Sequence):
		return retval
		retval = np.concatenate(retval, axis=-1)
		return undo_transforms(retval, transformers)


		class DAGModel(TensorGraph):

		@@ -877,8 +867,6 @@ class MPNNModel(TensorGraph):
		pad_batches=True):
		""" Same generator as Weave models """
		for epoch in range(epochs):
		if not predict:
		print('Starting epoch %i' % epoch)
		for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(
		batch_size=self.batch_size,
		deterministic=deterministic,

deepchem/models/tensorgraph/models/test_graph_models.py

+29 −2

Original line number	Diff line number	Diff line
		import unittest

		import os
		import numpy as np
		import scipy

		import deepchem as dc
		from deepchem.data import NumpyDataset
		@@ -193,7 +194,7 @@ class TestGraphModels(unittest.TestCase):

		model = WeaveModel(len(tasks), mode='regression')

		model.fit(dataset, nb_epoch=70)
		model.fit(dataset, nb_epoch=80)
		scores = model.evaluate(dataset, [metric], transformers)
		assert all(s < 0.1 for s in scores['mean_absolute_error'])

		@@ -203,6 +204,7 @@ class TestGraphModels(unittest.TestCase):
		assert np.allclose(scores['mean_absolute_error'],
		scores2['mean_absolute_error'])

		@attr("slow")
		def test_dag_model(self):
		tasks, dataset, transformers, metric = self.get_dataset(
		'classification', 'GraphConv')
		@@ -353,3 +355,28 @@ class TestGraphModels(unittest.TestCase):
		assert mean_error < 0.5 * mean_value
		assert mean_std > 0.5 * mean_error
		assert mean_std < mean_value

		def test_dtnn_regression_model(self):
		current_dir = os.path.dirname(os.path.abspath(__file__))
		input_file = os.path.join(current_dir, "example_DTNN.mat")
		dataset = scipy.io.loadmat(input_file)
		X = dataset['X']
		y = dataset['T']
		w = np.ones_like(y)
		dataset = dc.data.NumpyDataset(X, y, w, ids=None)
		n_tasks = y.shape[1]

		model = dc.models.DTNNModel(
		n_tasks,
		n_embedding=20,
		n_distance=100,
		learning_rate=1.0,
		mode="regression")

		# Fit trained model
		model.fit(dataset, nb_epoch=200)

		# Eval model on train
		pred = model.predict(dataset)
		mean_rel_error = np.mean(np.abs(1 - pred / y))
		assert mean_rel_error < 0.1

deepchem/models/tests/example_DTNN.mat

deleted100644 → 0

−13.4 KiB

File deleted.

View file

deepchem/models/tests/test_overfit.py

+0 −33

Original line number	Diff line number	Diff line
		@@ -524,39 +524,6 @@ class TestOverfit(test_util.TensorFlowTestCase):
		scores = model.evaluate(dataset, [regression_metric])
		assert scores[regression_metric.name] < .2

		def test_tensorgraph_DTNN_multitask_regression_overfit(self):
		"""Test deep tensor neural net overfits tiny data."""
		np.random.seed(123)
		tf.set_random_seed(123)

		input_file = os.path.join(self.current_dir, "example_DTNN.mat")
		dataset = scipy.io.loadmat(input_file)
		X = dataset['X']
		y = dataset['T']
		w = np.ones_like(y)
		dataset = dc.data.DiskDataset.from_numpy(X, y, w, ids=None)
		regression_metric = dc.metrics.Metric(
		dc.metrics.pearson_r2_score, task_averager=np.mean)
		n_tasks = y.shape[1]
		batch_size = 10

		model = dc.models.DTNNModel(
		n_tasks,
		n_embedding=20,
		n_distance=100,
		batch_size=batch_size,
		learning_rate=0.001,
		use_queue=False,
		mode="regression")

		# Fit trained model
		model.fit(dataset, nb_epoch=20)

		# Eval model on train
		scores = model.evaluate(dataset, [regression_metric])

		assert scores[regression_metric.name] > .9

		@attr('slow')
		def test_ANI_multitask_regression_overfit(self):
		"""Test ANI-1 regression overfits tiny data."""

Admin message