Loading README.md +282 −227 File changed.Preview size limit exceeded, changes collapsed. Show changes deepchem/models/tensorflow_models/fcnet.py +36 −7 Original line number Diff line number Diff line Loading @@ -14,6 +14,7 @@ import collections import deepchem as dc from deepchem.nn import model_ops from deepchem.utils.save import log from deepchem.metrics import to_one_hot, from_one_hot from deepchem.models.tensorflow_models import TensorflowGraph from deepchem.models.tensorflow_models import TensorflowGraphModel from deepchem.models.tensorflow_models import TensorflowClassifier Loading Loading @@ -72,8 +73,7 @@ class TensorGraphMultiTaskClassifier(TensorGraph): n_classes: int the number of classes """ super(TensorGraphMultiTaskClassifier, self).__init__( mode='classification', **kwargs) super(TensorGraphMultiTaskClassifier, self).__init__(**kwargs) self.n_tasks = n_tasks self.n_features = n_features self.n_classes = n_classes Loading Loading @@ -150,6 +150,36 @@ class TensorGraphMultiTaskClassifier(TensorGraph): feed_dict[self.task_weights[0]] = w_b yield feed_dict def predict_proba(self, dataset, transformers=[], outputs=None): return super(TensorGraphMultiTaskClassifier, self).predict( dataset, transformers, outputs) def predict(self, dataset, transformers=[], outputs=None): """ Uses self to make predictions on provided Dataset object. Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns ------- y_pred: numpy ndarray or list of numpy ndarrays """ # Results is of shape (n_samples, n_tasks, n_classes) retval = super(TensorGraphMultiTaskClassifier, self).predict( dataset, transformers, outputs) # retval is of shape (n_samples, n_tasks) return np.argmax(retval, axis=2) class TensorGraphMultiTaskRegressor(TensorGraph): Loading Loading @@ -197,8 +227,7 @@ class TensorGraphMultiTaskRegressor(TensorGraph): len(layer_sizes). Alternatively this may be a single value instead of a list, in which case the same value is used for every layer. """ super(TensorGraphMultiTaskRegressor, self).__init__( mode='regression', **kwargs) super(TensorGraphMultiTaskRegressor, self).__init__(**kwargs) self.n_tasks = n_tasks self.n_features = n_features n_layers = len(layer_sizes) Loading Loading @@ -362,7 +391,7 @@ class TensorGraphMultiTaskFitTransformRegressor(TensorGraphMultiTaskRegressor): feed_dict[self.task_weights[0]] = w_b yield feed_dict def predict_proba_on_generator(self, generator, transformers=[]): def predict_on_generator(self, generator, transformers=[], outputs=None): def transform_generator(): for feed_dict in generator: Loading @@ -375,8 +404,8 @@ class TensorGraphMultiTaskFitTransformRegressor(TensorGraphMultiTaskRegressor): yield feed_dict return super(TensorGraphMultiTaskFitTransformRegressor, self).predict_proba_on_generator(transform_generator(), transformers) self).predict_on_generator(transform_generator(), transformers, outputs) class TensorflowMultiTaskClassifier(TensorflowClassifier): Loading deepchem/models/tensorgraph/models/graph_models.py +79 −68 Original line number Diff line number Diff line Loading @@ -27,6 +27,7 @@ class WeaveTensorGraph(TensorGraph): n_pair_feat=14, n_hidden=50, n_graph_feat=128, mode="classification", **kwargs): """ Parameters Loading @@ -41,13 +42,15 @@ class WeaveTensorGraph(TensorGraph): Number of units(convolution depths) in corresponding hidden layer n_graph_feat: int, optional Number of output features for each molecule(graph) mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.n_atom_feat = n_atom_feat self.n_pair_feat = n_pair_feat self.n_hidden = n_hidden self.n_graph_feat = n_graph_feat self.mode = mode super(WeaveTensorGraph, self).__init__(**kwargs) self.build_graph() Loading Loading @@ -187,6 +190,7 @@ class DTNNTensorGraph(TensorGraph): distance_min=-1, distance_max=18, output_activation=True, mode="classification", **kwargs): """ Parameters Loading @@ -204,7 +208,8 @@ class DTNNTensorGraph(TensorGraph): minimum distance of atom pairs, default = -1 Angstorm distance_max: float, optional maximum distance of atom pairs, default = 18 Angstorm mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.n_embedding = n_embedding Loading @@ -217,6 +222,7 @@ class DTNNTensorGraph(TensorGraph): [distance_min + i * self.step_size for i in range(n_distance)]) self.steps = np.expand_dims(self.steps, 0) self.output_activation = output_activation self.mode = mode super(DTNNTensorGraph, self).__init__(**kwargs) assert self.mode == "regression" self.build_graph() Loading Loading @@ -335,6 +341,7 @@ class DAGTensorGraph(TensorGraph): n_atom_feat=75, n_graph_feat=30, n_outputs=30, mode="classification", **kwargs): """ Parameters Loading @@ -349,13 +356,15 @@ class DAGTensorGraph(TensorGraph): Number of features for atom in the graph n_outputs: int, optional Number of features for each molecule mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.max_atoms = max_atoms self.n_atom_feat = n_atom_feat self.n_graph_feat = n_graph_feat self.n_outputs = n_outputs self.mode = mode super(DAGTensorGraph, self).__init__(**kwargs) self.build_graph() Loading Loading @@ -477,15 +486,17 @@ class DAGTensorGraph(TensorGraph): class GraphConvTensorGraph(TensorGraph): def __init__(self, n_tasks, **kwargs): def __init__(self, n_tasks, mode="classification", **kwargs): """ Parameters ---------- n_tasks: int Number of tasks mode: str Either "classification" or "regression" """ self.n_tasks = n_tasks self.mode = mode self.error_bars = True if 'error_bars' in kwargs and kwargs['error_bars'] else False kwargs['use_queue'] = False super(GraphConvTensorGraph, self).__init__(**kwargs) Loading deepchem/models/tensorgraph/tensor_graph.py +109 −91 Original line number Diff line number Diff line Loading @@ -3,6 +3,7 @@ import threading import time import networkx as nx import collections import numpy as np import os import six Loading @@ -28,12 +29,10 @@ class TensorGraph(Model): batch_size=100, random_seed=None, use_queue=True, mode="regression", graph=None, learning_rate=0.001, **kwargs): """ TODO(LESWING) allow a model to change its learning rate Parameters ---------- tensorboard: bool Loading @@ -48,10 +47,6 @@ class TensorGraph(Model): queue in batches of self.batch_size in a separate thread from the thread training the model. You cannot use a queue when batches are not of consistent size mode: str "regression" or "classification". "classification" models on predict will do an argmax(axis=2) to determine the class of the prediction. graph: tensorflow.Graph the Graph in which to create Tensorflow objects. If None, a new Graph is created. Loading Loading @@ -85,7 +80,6 @@ class TensorGraph(Model): self.tensorboard = tensorboard self.tensorboard_log_frequency = tensorboard_log_frequency self.tensorboard_step = 0 self.mode = mode self.global_step = 0 self.last_checkpoint = None self.use_queue = use_queue Loading Loading @@ -185,10 +179,8 @@ class TensorGraph(Model): avg_loss)) saver.save(sess, self.save_file, global_step=self.global_step) self.last_checkpoint = saver.last_checkpoints[-1] ############################################################## TIMING time2 = time.time() print("TIMING: model fitting took %0.3f s" % (time2 - time1)) ############################################################## TIMING def _log_tensorboard(self, summary): """ Loading Loading @@ -238,123 +230,146 @@ class TensorGraph(Model): feed_dict[initial_state] = zero_state yield feed_dict def predict_on_generator(self, generator, transformers=[]): """Generates output predictions for the input samples, processing the samples in a batched way. # Arguments x: the input data, as a Numpy array. batch_size: integer. verbose: verbosity mode, 0 or 1. # Returns A Numpy array of predictions. """ retval = self.predict_proba_on_generator(generator, transformers) if self.mode == 'classification': retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1) return retval def predict_proba_on_generator(self, generator, transformers=[]): def predict_on_generator(self, generator, transformers=[], outputs=None): """ Parameters ---------- generator: Generator Generator that constructs feed dictionaries for TensorGraph. transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs. If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns: y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) """ if not self.built: self.build() if outputs is None: outputs = self.outputs elif not isinstance(outputs, collections.Sequence): outputs = [outputs] with self._get_tf("Graph").as_default(): with tf.Session() as sess: saver = tf.train.Saver() self._initialize_weights(sess, saver) out_tensors = [x.out_tensor for x in self.outputs] results = [] # Gather results for each output results = [[] for out in out_tensors] for feed_dict in generator: feed_dict = { self.layers[k.name].out_tensor: v for k, v in six.iteritems(feed_dict) } feed_dict[self._training_placeholder] = 0.0 result = np.array(sess.run(out_tensors, feed_dict=feed_dict)) if len(result.shape) == 3: result = np.transpose(result, axes=[1, 0, 2]) result = undo_transforms(result, transformers) results.append(result) return np.concatenate(results, axis=0) def bayesian_predict_on_batch(self, X, transformers=[], n_passes=4): feed_results = sess.run(out_tensors, feed_dict=feed_dict) if len(feed_results) > 1: if len(transformers): raise ValueError("Does not support transformations " "for multiple outputs.") elif len(feed_results) == 1: result = undo_transforms(feed_results[0], transformers) feed_results = [result] for ind, result in enumerate(feed_results): results[ind].append(result) final_results = [] for result_list in results: final_results.append(np.concatenate(result_list, axis=0)) # If only one output, just return array if len(final_results) == 1: return final_results[0] else: return final_results def predict_proba_on_generator(self, generator, transformers=[], outputs=None): """ Returns: mu: numpy ndarray of shape (n_samples, n_tasks) sigma: numpy ndarray of shape (n_samples, n_tasks) """ dataset = NumpyDataset(X=X, y=None, n_tasks=len(self.outputs)) y_ = [] for i in range(n_passes): generator = self.default_generator( dataset, predict=True, pad_batches=True) y_.append(self.predict_on_generator(generator, transformers)) y_ = np.concatenate(y_, axis=2) mu = np.mean(y_, axis=2) sigma = np.std(y_, axis=2) return mu, sigma def predict_on_smiles_batch(self, smiles, featurizer, n_tasks, transformers=[]): """ # Returns: A numpy ndarray of shape (n_samples, n_tasks) y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) """ convmols = featurize_smiles_np(smiles, featurizer) dataset = NumpyDataset(X=convmols, y=None, n_tasks=len(self.outputs)) generator = self.default_generator(dataset, predict=True, pad_batches=True) return self.predict_on_generator(generator, transformers) return self.predict_on_generator(generator, transformers, outputs) def predict_on_batch(self, X, sess=None, transformers=[]): """Generates output predictions for the input samples, processing the samples in a batched way. def predict_on_batch(self, X, transformers=[]): """Generates predictions for input samples, processing samples in a batch. # Arguments x: the input data, as a Numpy array. batch_size: integer. verbose: verbosity mode, 0 or 1. Parameters ---------- X: ndarray the input data, as a Numpy array. transformers: List List of dc.trans.Transformers # Returns Returns ------- A Numpy array of predictions. """ dataset = NumpyDataset(X=X, y=None) generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_on_generator(generator, transformers) def predict_proba_on_batch(self, X, sess=None, transformers=[]): dataset = NumpyDataset(X=X, y=None) generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_proba_on_generator(generator, transformers) def predict_proba_on_batch(self, X, transformers=[]): """Generates predictions for input samples, processing samples in a batch. Parameters ---------- X: ndarray the input data, as a Numpy array. transformers: List List of dc.trans.Transformers Returns ------- A Numpy array of predictions. """ return self.predict_on_batch(X, transformers) def predict(self, dataset, transformers=[], batch_size=None): def predict(self, dataset, transformers=[], outputs=None): """ Uses self to make predictions on provided Dataset object. Returns: y_pred: numpy ndarray of shape (n_samples,) Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns ------- results: numpy ndarray or list of numpy ndarrays """ generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_on_generator(generator, transformers) return self.predict_on_generator(generator, transformers, outputs) def predict_proba(self, dataset, transformers=[], batch_size=None): def predict_proba(self, dataset, transformers=[], outputs=None): """ TODO: Do transformers even make sense here? Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns: y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) Returns ------- y_pred: numpy ndarray or list of numpy ndarrays """ generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_proba_on_generator(generator, transformers) return self.predict_proba_on_generator(generator, transformers, outputs) def topsort(self): return nx.topological_sort(self.nxgraph) Loading Loading @@ -550,12 +565,15 @@ class TensorGraph(Model): return self._get_tf("GlobalStep") def _get_tf(self, obj): """ TODO(LESWING) REALLY NEED TO DOCUMENT THIS """Fetches underlying TensorFlow primitives. Parameters ---------- obj obj: str If "Graph", returns tf.Graph instance. If "FileWriter", returns tf.summary.FileWriter. If "Optimizer", returns the optimizer. If "train_op", returns the train operation. If "summary_op", returns the merged summary. If "GlobalStep" returns the global step. Returns ------- TensorFlow Object Loading deepchem/models/tensorgraph/tests/test_tensor_graph.py +13 −14 Original line number Diff line number Diff line Loading @@ -37,7 +37,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=0.4))) @flaky Loading Loading @@ -78,10 +78,10 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1000, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) predictions = tg.predict_on_generator(databag.iterbatches()) for i in range(2): y_real = ys[i].X y_pred = prediction[:, i, :] y_pred = predictions[i] assert_true(np.all(np.isclose(y_pred, y_real, atol=0.6))) def test_single_task_regressor(self): Loading @@ -98,7 +98,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(dense) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=3.0))) def test_multi_task_regressor(self): Loading Loading @@ -137,10 +137,10 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1000, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) predictions = tg.predict_on_generator(databag.iterbatches()) for i in range(2): y_real = ys[i].X y_pred = prediction[:, i, :] y_pred = predictions[i] assert_true(np.all(np.isclose(y_pred, y_real, atol=1.5))) @flaky Loading @@ -160,7 +160,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=0.4))) @flaky Loading @@ -184,11 +184,11 @@ class TestTensorGraph(unittest.TestCase): initial_rate=0.1, decay_rate=0.96, decay_steps=100000) tg.set_optimizer(GradientDescent(learning_rate=learning_rate)) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) tg.save() tg1 = TensorGraph.load_from_dir(tg.model_dir) prediction2 = np.squeeze(tg1.predict_proba_on_batch(X)) prediction2 = np.squeeze(tg1.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, prediction2, atol=0.01))) @nottest Loading Loading @@ -235,11 +235,11 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) tg.save() tg1 = TensorGraph.load_from_dir(tg.model_dir) prediction2 = np.squeeze(tg1.predict_proba_on_batch(X)) prediction2 = np.squeeze(tg1.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, prediction2, atol=0.01))) def test_shared_layer(self): Loading Loading @@ -279,6 +279,5 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) assert_true( np.all(np.isclose(prediction[:, 0], prediction[:, 1], atol=0.01))) prediction = tg.predict_on_generator(databag.iterbatches()) assert_true(np.all(np.isclose(prediction[0], prediction[1], atol=0.01))) Loading
deepchem/models/tensorflow_models/fcnet.py +36 −7 Original line number Diff line number Diff line Loading @@ -14,6 +14,7 @@ import collections import deepchem as dc from deepchem.nn import model_ops from deepchem.utils.save import log from deepchem.metrics import to_one_hot, from_one_hot from deepchem.models.tensorflow_models import TensorflowGraph from deepchem.models.tensorflow_models import TensorflowGraphModel from deepchem.models.tensorflow_models import TensorflowClassifier Loading Loading @@ -72,8 +73,7 @@ class TensorGraphMultiTaskClassifier(TensorGraph): n_classes: int the number of classes """ super(TensorGraphMultiTaskClassifier, self).__init__( mode='classification', **kwargs) super(TensorGraphMultiTaskClassifier, self).__init__(**kwargs) self.n_tasks = n_tasks self.n_features = n_features self.n_classes = n_classes Loading Loading @@ -150,6 +150,36 @@ class TensorGraphMultiTaskClassifier(TensorGraph): feed_dict[self.task_weights[0]] = w_b yield feed_dict def predict_proba(self, dataset, transformers=[], outputs=None): return super(TensorGraphMultiTaskClassifier, self).predict( dataset, transformers, outputs) def predict(self, dataset, transformers=[], outputs=None): """ Uses self to make predictions on provided Dataset object. Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns ------- y_pred: numpy ndarray or list of numpy ndarrays """ # Results is of shape (n_samples, n_tasks, n_classes) retval = super(TensorGraphMultiTaskClassifier, self).predict( dataset, transformers, outputs) # retval is of shape (n_samples, n_tasks) return np.argmax(retval, axis=2) class TensorGraphMultiTaskRegressor(TensorGraph): Loading Loading @@ -197,8 +227,7 @@ class TensorGraphMultiTaskRegressor(TensorGraph): len(layer_sizes). Alternatively this may be a single value instead of a list, in which case the same value is used for every layer. """ super(TensorGraphMultiTaskRegressor, self).__init__( mode='regression', **kwargs) super(TensorGraphMultiTaskRegressor, self).__init__(**kwargs) self.n_tasks = n_tasks self.n_features = n_features n_layers = len(layer_sizes) Loading Loading @@ -362,7 +391,7 @@ class TensorGraphMultiTaskFitTransformRegressor(TensorGraphMultiTaskRegressor): feed_dict[self.task_weights[0]] = w_b yield feed_dict def predict_proba_on_generator(self, generator, transformers=[]): def predict_on_generator(self, generator, transformers=[], outputs=None): def transform_generator(): for feed_dict in generator: Loading @@ -375,8 +404,8 @@ class TensorGraphMultiTaskFitTransformRegressor(TensorGraphMultiTaskRegressor): yield feed_dict return super(TensorGraphMultiTaskFitTransformRegressor, self).predict_proba_on_generator(transform_generator(), transformers) self).predict_on_generator(transform_generator(), transformers, outputs) class TensorflowMultiTaskClassifier(TensorflowClassifier): Loading
deepchem/models/tensorgraph/models/graph_models.py +79 −68 Original line number Diff line number Diff line Loading @@ -27,6 +27,7 @@ class WeaveTensorGraph(TensorGraph): n_pair_feat=14, n_hidden=50, n_graph_feat=128, mode="classification", **kwargs): """ Parameters Loading @@ -41,13 +42,15 @@ class WeaveTensorGraph(TensorGraph): Number of units(convolution depths) in corresponding hidden layer n_graph_feat: int, optional Number of output features for each molecule(graph) mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.n_atom_feat = n_atom_feat self.n_pair_feat = n_pair_feat self.n_hidden = n_hidden self.n_graph_feat = n_graph_feat self.mode = mode super(WeaveTensorGraph, self).__init__(**kwargs) self.build_graph() Loading Loading @@ -187,6 +190,7 @@ class DTNNTensorGraph(TensorGraph): distance_min=-1, distance_max=18, output_activation=True, mode="classification", **kwargs): """ Parameters Loading @@ -204,7 +208,8 @@ class DTNNTensorGraph(TensorGraph): minimum distance of atom pairs, default = -1 Angstorm distance_max: float, optional maximum distance of atom pairs, default = 18 Angstorm mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.n_embedding = n_embedding Loading @@ -217,6 +222,7 @@ class DTNNTensorGraph(TensorGraph): [distance_min + i * self.step_size for i in range(n_distance)]) self.steps = np.expand_dims(self.steps, 0) self.output_activation = output_activation self.mode = mode super(DTNNTensorGraph, self).__init__(**kwargs) assert self.mode == "regression" self.build_graph() Loading Loading @@ -335,6 +341,7 @@ class DAGTensorGraph(TensorGraph): n_atom_feat=75, n_graph_feat=30, n_outputs=30, mode="classification", **kwargs): """ Parameters Loading @@ -349,13 +356,15 @@ class DAGTensorGraph(TensorGraph): Number of features for atom in the graph n_outputs: int, optional Number of features for each molecule mode: str Either "classification" or "regression" for type of model. """ self.n_tasks = n_tasks self.max_atoms = max_atoms self.n_atom_feat = n_atom_feat self.n_graph_feat = n_graph_feat self.n_outputs = n_outputs self.mode = mode super(DAGTensorGraph, self).__init__(**kwargs) self.build_graph() Loading Loading @@ -477,15 +486,17 @@ class DAGTensorGraph(TensorGraph): class GraphConvTensorGraph(TensorGraph): def __init__(self, n_tasks, **kwargs): def __init__(self, n_tasks, mode="classification", **kwargs): """ Parameters ---------- n_tasks: int Number of tasks mode: str Either "classification" or "regression" """ self.n_tasks = n_tasks self.mode = mode self.error_bars = True if 'error_bars' in kwargs and kwargs['error_bars'] else False kwargs['use_queue'] = False super(GraphConvTensorGraph, self).__init__(**kwargs) Loading
deepchem/models/tensorgraph/tensor_graph.py +109 −91 Original line number Diff line number Diff line Loading @@ -3,6 +3,7 @@ import threading import time import networkx as nx import collections import numpy as np import os import six Loading @@ -28,12 +29,10 @@ class TensorGraph(Model): batch_size=100, random_seed=None, use_queue=True, mode="regression", graph=None, learning_rate=0.001, **kwargs): """ TODO(LESWING) allow a model to change its learning rate Parameters ---------- tensorboard: bool Loading @@ -48,10 +47,6 @@ class TensorGraph(Model): queue in batches of self.batch_size in a separate thread from the thread training the model. You cannot use a queue when batches are not of consistent size mode: str "regression" or "classification". "classification" models on predict will do an argmax(axis=2) to determine the class of the prediction. graph: tensorflow.Graph the Graph in which to create Tensorflow objects. If None, a new Graph is created. Loading Loading @@ -85,7 +80,6 @@ class TensorGraph(Model): self.tensorboard = tensorboard self.tensorboard_log_frequency = tensorboard_log_frequency self.tensorboard_step = 0 self.mode = mode self.global_step = 0 self.last_checkpoint = None self.use_queue = use_queue Loading Loading @@ -185,10 +179,8 @@ class TensorGraph(Model): avg_loss)) saver.save(sess, self.save_file, global_step=self.global_step) self.last_checkpoint = saver.last_checkpoints[-1] ############################################################## TIMING time2 = time.time() print("TIMING: model fitting took %0.3f s" % (time2 - time1)) ############################################################## TIMING def _log_tensorboard(self, summary): """ Loading Loading @@ -238,123 +230,146 @@ class TensorGraph(Model): feed_dict[initial_state] = zero_state yield feed_dict def predict_on_generator(self, generator, transformers=[]): """Generates output predictions for the input samples, processing the samples in a batched way. # Arguments x: the input data, as a Numpy array. batch_size: integer. verbose: verbosity mode, 0 or 1. # Returns A Numpy array of predictions. """ retval = self.predict_proba_on_generator(generator, transformers) if self.mode == 'classification': retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1) return retval def predict_proba_on_generator(self, generator, transformers=[]): def predict_on_generator(self, generator, transformers=[], outputs=None): """ Parameters ---------- generator: Generator Generator that constructs feed dictionaries for TensorGraph. transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs. If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns: y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) """ if not self.built: self.build() if outputs is None: outputs = self.outputs elif not isinstance(outputs, collections.Sequence): outputs = [outputs] with self._get_tf("Graph").as_default(): with tf.Session() as sess: saver = tf.train.Saver() self._initialize_weights(sess, saver) out_tensors = [x.out_tensor for x in self.outputs] results = [] # Gather results for each output results = [[] for out in out_tensors] for feed_dict in generator: feed_dict = { self.layers[k.name].out_tensor: v for k, v in six.iteritems(feed_dict) } feed_dict[self._training_placeholder] = 0.0 result = np.array(sess.run(out_tensors, feed_dict=feed_dict)) if len(result.shape) == 3: result = np.transpose(result, axes=[1, 0, 2]) result = undo_transforms(result, transformers) results.append(result) return np.concatenate(results, axis=0) def bayesian_predict_on_batch(self, X, transformers=[], n_passes=4): feed_results = sess.run(out_tensors, feed_dict=feed_dict) if len(feed_results) > 1: if len(transformers): raise ValueError("Does not support transformations " "for multiple outputs.") elif len(feed_results) == 1: result = undo_transforms(feed_results[0], transformers) feed_results = [result] for ind, result in enumerate(feed_results): results[ind].append(result) final_results = [] for result_list in results: final_results.append(np.concatenate(result_list, axis=0)) # If only one output, just return array if len(final_results) == 1: return final_results[0] else: return final_results def predict_proba_on_generator(self, generator, transformers=[], outputs=None): """ Returns: mu: numpy ndarray of shape (n_samples, n_tasks) sigma: numpy ndarray of shape (n_samples, n_tasks) """ dataset = NumpyDataset(X=X, y=None, n_tasks=len(self.outputs)) y_ = [] for i in range(n_passes): generator = self.default_generator( dataset, predict=True, pad_batches=True) y_.append(self.predict_on_generator(generator, transformers)) y_ = np.concatenate(y_, axis=2) mu = np.mean(y_, axis=2) sigma = np.std(y_, axis=2) return mu, sigma def predict_on_smiles_batch(self, smiles, featurizer, n_tasks, transformers=[]): """ # Returns: A numpy ndarray of shape (n_samples, n_tasks) y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) """ convmols = featurize_smiles_np(smiles, featurizer) dataset = NumpyDataset(X=convmols, y=None, n_tasks=len(self.outputs)) generator = self.default_generator(dataset, predict=True, pad_batches=True) return self.predict_on_generator(generator, transformers) return self.predict_on_generator(generator, transformers, outputs) def predict_on_batch(self, X, sess=None, transformers=[]): """Generates output predictions for the input samples, processing the samples in a batched way. def predict_on_batch(self, X, transformers=[]): """Generates predictions for input samples, processing samples in a batch. # Arguments x: the input data, as a Numpy array. batch_size: integer. verbose: verbosity mode, 0 or 1. Parameters ---------- X: ndarray the input data, as a Numpy array. transformers: List List of dc.trans.Transformers # Returns Returns ------- A Numpy array of predictions. """ dataset = NumpyDataset(X=X, y=None) generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_on_generator(generator, transformers) def predict_proba_on_batch(self, X, sess=None, transformers=[]): dataset = NumpyDataset(X=X, y=None) generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_proba_on_generator(generator, transformers) def predict_proba_on_batch(self, X, transformers=[]): """Generates predictions for input samples, processing samples in a batch. Parameters ---------- X: ndarray the input data, as a Numpy array. transformers: List List of dc.trans.Transformers Returns ------- A Numpy array of predictions. """ return self.predict_on_batch(X, transformers) def predict(self, dataset, transformers=[], batch_size=None): def predict(self, dataset, transformers=[], outputs=None): """ Uses self to make predictions on provided Dataset object. Returns: y_pred: numpy ndarray of shape (n_samples,) Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns ------- results: numpy ndarray or list of numpy ndarrays """ generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_on_generator(generator, transformers) return self.predict_on_generator(generator, transformers, outputs) def predict_proba(self, dataset, transformers=[], batch_size=None): def predict_proba(self, dataset, transformers=[], outputs=None): """ TODO: Do transformers even make sense here? Parameters ---------- dataset: dc.data.Dataset Dataset to make prediction on transformers: list List of dc.trans.Transformers. outputs: object If outputs is None, then will assume outputs = self.outputs[0] (single output). If outputs is a Layer/Tensor, then will evaluate and return as a single ndarray. If outputs is a list of Layers/Tensors, will return a list of ndarrays. Returns: y_pred: numpy ndarray of shape (n_samples, n_classes*n_tasks) Returns ------- y_pred: numpy ndarray or list of numpy ndarrays """ generator = self.default_generator(dataset, predict=True, pad_batches=False) return self.predict_proba_on_generator(generator, transformers) return self.predict_proba_on_generator(generator, transformers, outputs) def topsort(self): return nx.topological_sort(self.nxgraph) Loading Loading @@ -550,12 +565,15 @@ class TensorGraph(Model): return self._get_tf("GlobalStep") def _get_tf(self, obj): """ TODO(LESWING) REALLY NEED TO DOCUMENT THIS """Fetches underlying TensorFlow primitives. Parameters ---------- obj obj: str If "Graph", returns tf.Graph instance. If "FileWriter", returns tf.summary.FileWriter. If "Optimizer", returns the optimizer. If "train_op", returns the train operation. If "summary_op", returns the merged summary. If "GlobalStep" returns the global step. Returns ------- TensorFlow Object Loading
deepchem/models/tensorgraph/tests/test_tensor_graph.py +13 −14 Original line number Diff line number Diff line Loading @@ -37,7 +37,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=0.4))) @flaky Loading Loading @@ -78,10 +78,10 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1000, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) predictions = tg.predict_on_generator(databag.iterbatches()) for i in range(2): y_real = ys[i].X y_pred = prediction[:, i, :] y_pred = predictions[i] assert_true(np.all(np.isclose(y_pred, y_real, atol=0.6))) def test_single_task_regressor(self): Loading @@ -98,7 +98,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(dense) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=3.0))) def test_multi_task_regressor(self): Loading Loading @@ -137,10 +137,10 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1000, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) predictions = tg.predict_on_generator(databag.iterbatches()) for i in range(2): y_real = ys[i].X y_pred = prediction[:, i, :] y_pred = predictions[i] assert_true(np.all(np.isclose(y_pred, y_real, atol=1.5))) @flaky Loading @@ -160,7 +160,7 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, y, atol=0.4))) @flaky Loading @@ -184,11 +184,11 @@ class TestTensorGraph(unittest.TestCase): initial_rate=0.1, decay_rate=0.96, decay_steps=100000) tg.set_optimizer(GradientDescent(learning_rate=learning_rate)) tg.fit(dataset, nb_epoch=1000) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) tg.save() tg1 = TensorGraph.load_from_dir(tg.model_dir) prediction2 = np.squeeze(tg1.predict_proba_on_batch(X)) prediction2 = np.squeeze(tg1.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, prediction2, atol=0.01))) @nottest Loading Loading @@ -235,11 +235,11 @@ class TestTensorGraph(unittest.TestCase): tg.add_output(output) tg.set_loss(loss) tg.fit(dataset, nb_epoch=1) prediction = np.squeeze(tg.predict_proba_on_batch(X)) prediction = np.squeeze(tg.predict_on_batch(X)) tg.save() tg1 = TensorGraph.load_from_dir(tg.model_dir) prediction2 = np.squeeze(tg1.predict_proba_on_batch(X)) prediction2 = np.squeeze(tg1.predict_on_batch(X)) assert_true(np.all(np.isclose(prediction, prediction2, atol=0.01))) def test_shared_layer(self): Loading Loading @@ -279,6 +279,5 @@ class TestTensorGraph(unittest.TestCase): tg.fit_generator( databag.iterbatches( epochs=1, batch_size=tg.batch_size, pad_batches=True)) prediction = tg.predict_proba_on_generator(databag.iterbatches()) assert_true( np.all(np.isclose(prediction[:, 0], prediction[:, 1], atol=0.01))) prediction = tg.predict_on_generator(databag.iterbatches()) assert_true(np.all(np.isclose(prediction[0], prediction[1], atol=0.01)))
