BPSymmetryFunction change (bb86fa8f) · Commits · 钟慕尧 / deepchem

deepchem/models/tensorgraph/models/symmetry_function_regression.py

+18 −37

Original line number	Diff line number	Diff line
		@@ -19,7 +19,7 @@ from deepchem.models.tensorgraph.symmetry_functions import DistanceMatrix, \

		class BPSymmetryFunctionRegression(TensorGraph):

		def __init__(self, n_tasks, max_atoms, n_hidden=40, n_embedding=10, **kwargs):
		def __init__(self, n_tasks, max_atoms, n_feat=96, layer_structures=[128, 64], **kwargs):
		"""
		Parameters
		----------
		@@ -32,49 +32,30 @@ class BPSymmetryFunctionRegression(TensorGraph):
		"""
		self.n_tasks = n_tasks
		self.max_atoms = max_atoms
		self.n_hidden = n_hidden
		self.n_embedding = n_embedding
		self.n_feat = n_feat
		self.layer_structures = layer_structures
		super(BPSymmetryFunctionRegression, self).__init__(**kwargs)
		self.build_graph()

		def build_graph(self):
		self.atom_numbers = Feature(shape=(None, self.max_atoms), dtype=tf.int32)
		self.atom_flags = Feature(shape=(None, self.max_atoms, self.max_atoms))
		self.atom_coordinates = Feature(shape=(None, self.max_atoms, 3))

		distance_matrix = DistanceMatrix(
		self.max_atoms, in_layers=[self.atom_coordinates, self.atom_flags])
		distance_cutoff = DistanceCutoff(
		self.max_atoms,
		cutoff=6 / 0.52917721092,
		in_layers=[distance_matrix, self.atom_flags])
		radial_symmetry = RadialSymmetry(
		self.max_atoms, in_layers=[distance_cutoff, distance_matrix])
		angular_symmetry = AngularSymmetry(
		self.max_atoms,
		in_layers=[distance_cutoff, distance_matrix, self.atom_coordinates])
		atom_embedding = DTNNEmbedding(
		n_embedding=self.n_embedding, in_layers=[self.atom_numbers])

		feature_merge = BPFeatureMerge(
		self.max_atoms,
		in_layers=[
		atom_embedding, radial_symmetry, angular_symmetry, self.atom_flags
		])
		self.atom_feats = Feature(shape=(None, self.max_atoms, self.n_feat))
		previous_layer = self.atom_feats

		Hiddens = []
		for n_hidden in self.layer_structures:
		Hidden = Dense(
		out_channels=self.n_hidden,
		activation_fn=tf.nn.tanh,
		in_layers=[feature_merge])
		Hidden2 = Dense(
		out_channels=self.n_hidden,
		out_channels=n_hidden,
		activation_fn=tf.nn.tanh,
		in_layers=[Hidden])
		in_layers=[previous_layer])
		Hiddens.append(Hidden)
		previous_layer = Hiddens[-1]

		costs = []
		self.labels_fd = []
		for task in range(self.n_tasks):
		regression = Dense(
		out_channels=1, activation_fn=None, in_layers=[Hidden2])
		out_channels=1, activation_fn=None, in_layers=[Hiddens[-1]])
		output = BPGather(self.max_atoms, in_layers=[regression, self.atom_flags])
		self.add_output(output)

		@@ -88,6 +69,7 @@ class BPSymmetryFunctionRegression(TensorGraph):
		loss = WeightedError(in_layers=[all_cost, self.weights])
		self.set_loss(loss)


		def default_generator(self,
		dataset,
		epochs=1,
		@@ -111,8 +93,7 @@ class BPSymmetryFunctionRegression(TensorGraph):
		flags = np.sign(np.array(X_b[:, :, 0]))
		feed_dict[self.atom_flags] = np.stack([flags]self.max_atoms, axis=2)\
		np.stack([flags]*self.max_atoms, axis=1)
		feed_dict[self.atom_numbers] = np.array(X_b[:, :, 0], dtype=int)
		feed_dict[self.atom_coordinates] = np.array(X_b[:, :, 1:], dtype=float)
		feed_dict[self.atom_feats] = np.array(X_b[:, :, 1:], dtype=float)
		yield feed_dict

		class ANIRegression(TensorGraph):

deepchem/trans/transformers.py

+26 −18

Original line number	Diff line number	Diff line
		@@ -937,6 +937,7 @@ class ANITransformer(Transformer):
		radial_length=32,
		angular_length=8,
		atom_cases=[1, 6, 7, 8, 16],
		atomic_number_differentiated=True,
		coordinates_in_bohr=True,
		transform_X=True,
		transform_y=False,
		@@ -950,6 +951,7 @@ class ANITransformer(Transformer):
		self.radial_length = radial_length
		self.angular_length = angular_length
		self.atom_cases = atom_cases
		self.atomic_number_differentiated = atomic_number_differentiated
		self.coordinates_in_bohr = coordinates_in_bohr
		self.transform_X = transform_X
		self.transform_y = transform_y
		@@ -1039,6 +1041,7 @@ class ANITransformer(Transformer):
		ita = tf.to_float(np.reshape(ita, (1, 1, 1, -1)))

		out = tf.exp(-ita * tf.square(d - Rs)) * d_cutoff
		if self.atomic_number_differentiated:
		out_tensors = []
		for atom_type in self.atom_cases:
		selected_atoms = tf.expand_dims(
		@@ -1046,6 +1049,8 @@ class ANITransformer(Transformer):
		axis=3)
		out_tensors.append(tf.reduce_sum(out * selected_atoms, axis=2))
		return tf.concat(out_tensors, axis=2)
		else:
		return tf.reduce_sum(out, axis=2)


		def angular_symmetry(self, d_cutoff, d, atom_numbers, coordinates):
		@@ -1088,6 +1093,7 @@ class ANITransformer(Transformer):
		out_tensor = tf.pow((1. + tf.cos(theta - thetas))/2., zeta) * \
		tf.exp(-ita * tf.square((R_ij + R_ik)/2. - Rs)) * f_R_ij * f_R_ik

		if self.atomic_number_differentiated:
		out_tensors = []
		for id_j, atom_type_j in enumerate(self.atom_cases):
		for atom_type_k in self.atom_cases[id_j:]:
		@@ -1098,6 +1104,8 @@ class ANITransformer(Transformer):
		out_tensors.append(
		tf.reduce_sum(out_tensor * selected_atoms, axis=(2, 3)))
		return tf.concat(out_tensors, axis=2)
		else:
		return tf.reduce_sum(out_tensor, axis=(2, 3))

		def get_num_feats(self):
		n_feat = self.outputs.get_shape().as_list()[-1]

examples/qm7/qm7_BPSymmetryFunction.py

+14 −8

Original line number	Diff line number	Diff line
		@@ -14,23 +14,29 @@ tasks, datasets, transformers = dc.molnet.load_qm7_from_mat(
		featurizer='BPSymmetryFunction')
		train_dataset, valid_dataset, test_dataset = datasets

		# Batch size of models
		max_atoms = 23
		batch_size = 16
		layer_structures = [128, 128, 64]

		ANItransformer = dc.trans.ANITransformer(max_atoms=max_atoms,
		atomic_number_differentiated=False)
		train_dataset = ANItransformer.transform(train_dataset)
		valid_dataset = ANItransformer.transform(valid_dataset)
		test_dataset = ANItransformer.transform(test_dataset)
		n_feat = ANItransformer.get_num_feats() - 1

		# Fit models
		metric = [
		dc.metrics.Metric(dc.metrics.mean_absolute_error, mode="regression"),
		dc.metrics.Metric(dc.metrics.pearson_r2_score, mode="regression")
		]

		# Batch size of models
		max_atoms = 23
		n_hidden = 40
		n_embedding = 0
		batch_size = 16

		model = dc.models.BPSymmetryFunctionRegression(
		len(tasks),
		max_atoms,
		n_hidden=n_hidden,
		n_embedding=n_embedding,
		n_feat,
		layer_structures=layer_structures,
		batch_size=batch_size,
		learning_rate=0.001,
		use_queue=False,

Admin message