def create_and_eval_model(train_dataset, test_dataset, task_type,
model_params, model_name, model_dir, tasks):
"""Helper method to create model for test."""
# Fit model
task_types = {task: task_type for task in tasks}
model_params["data_shape"] = train_dataset.get_data_shape()
print("Creating Model object.")
import deepchem.models.deep
model = Model.model_builder(model_name, task_types, model_params)
print("About to fit model")
model.fit(train_dataset)
print("Done fitting, about to save...")
model.save(model_dir)
# Eval model on train
evaluator = Evaluator(model, train_dataset, verbose=True)
with tempfile.NamedTemporaryFile() as train_csv_out:
with tempfile.NamedTemporaryFile() as train_stats_out:
_, performance_df = evaluator.compute_model_performance(
train_csv_out, train_stats_out)
print("train_performance_df")
print(performance_df)
evaluator = Evaluator(model, test_dataset, verbose=True)
with tempfile.NamedTemporaryFile() as test_csv_out:
with tempfile.NamedTemporaryFile() as test_stats_out:
_, performance_df = evaluator.compute_model_performance(
test_csv_out, test_stats_out)
print("test_performance_df")
print(performance_df)
return performance_df.iterrows().next()[1]["r2_score"]
def test_sklearn_regression(self):
"""Test that sklearn models can learn on simple regression datasets."""
np.random.seed(123)
dataset = sklearn.datasets.load_diabetes()
X, y = dataset.data, dataset.target
frac_train = .7
n_samples = len(X)
X_train, y_train = X[:frac_train * n_samples], y[:frac_train *
n_samples]
X_test, y_test = X[frac_train * n_samples:], y[frac_train * n_samples:]
train_dataset = Dataset.from_numpy(self.train_dir, X_train, y_train)
test_dataset = Dataset.from_numpy(self.test_dir, X_test, y_test)
tasks = train_dataset.get_task_names()
task_types = {task: "regression" for task in tasks}
model_params = {
"batch_size": None,
"data_shape": train_dataset.get_data_shape()
}
verbosity = "high"
regression_metric = Metric(metrics.r2_score, verbosity=verbosity)
model = SklearnModel(tasks,
task_types,
model_params,
self.model_dir,
mode="regression",
model_instance=LinearRegression())
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
transformers = []
train_evaluator = Evaluator(model,
train_dataset,
transformers,
verbosity=verbosity)
train_scores = train_evaluator.compute_model_performance(
[regression_metric])
print("train_scores")
print(train_scores)
# Eval model on test
transformers = []
evaluator = Evaluator(model,
test_dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
print("scores")
print(scores)
assert scores[regression_metric.name] > .5
def test_singletask_sklearn_rf_ECFP_regression_API(self):
"""Test of singletask RF ECFP regression API."""
splittype = "scaffold"
featurizer = CircularFingerprint(size=1024)
model_params = {}
tasks = ["log-solubility"]
task_type = "regression"
task_types = {task: task_type for task in tasks}
input_file = os.path.join(self.current_dir, "example.csv")
loader = DataLoader(tasks=tasks,
smiles_field=self.smiles_field,
featurizer=featurizer,
verbosity="low")
dataset = loader.featurize(input_file, self.data_dir)
splitter = ScaffoldSplitter()
train_dataset, test_dataset = splitter.train_test_split(
dataset, self.train_dir, self.test_dir)
input_transformers = []
output_transformers = [
NormalizationTransformer(transform_y=True, dataset=train_dataset)
]
transformers = input_transformers + output_transformers
model_params["data_shape"] = train_dataset.get_data_shape()
regression_metrics = [
Metric(metrics.r2_score),
Metric(metrics.mean_squared_error),
Metric(metrics.mean_absolute_error)
]
model = SklearnModel(tasks,
task_types,
model_params,
self.model_dir,
mode="regression",
model_instance=RandomForestRegressor())
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
evaluator = Evaluator(model,
train_dataset,
transformers,
verbosity=True)
_ = evaluator.compute_model_performance(regression_metrics)
# Eval model on test
evaluator = Evaluator(model,
test_dataset,
transformers,
verbosity=True)
_ = evaluator.compute_model_performance(regression_metrics)
def test_sklearn_multitask_classification(self):
"""Test that sklearn models can learn on simple multitask classification."""
np.random.seed(123)
n_tasks = 4
tasks = range(n_tasks)
dataset = sklearn.datasets.load_digits(n_class=2)
X, y = dataset.data, dataset.target
y = np.reshape(y, (len(y), 1))
y = np.hstack([y] * n_tasks)
frac_train = .7
n_samples = len(X)
n_train = int(frac_train * n_samples)
X_train, y_train = X[:n_train], y[:n_train]
X_test, y_test = X[n_train:], y[n_train:]
train_dataset = DiskDataset.from_numpy(self.train_dir, X_train,
y_train)
test_dataset = DiskDataset.from_numpy(self.test_dir, X_test, y_test)
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
def model_builder(model_dir):
sklearn_model = LogisticRegression()
return SklearnModel(sklearn_model, model_dir)
model = SingletaskToMultitask(tasks, model_builder, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
transformers = []
train_evaluator = Evaluator(model,
train_dataset,
transformers,
verbosity=verbosity)
train_scores = train_evaluator.compute_model_performance(
[classification_metric])
# Eval model on test
transformers = []
evaluator = Evaluator(model,
test_dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
for score in scores[classification_metric.name]:
assert score > .5
def test_singletask_to_multitask_classification(self):
n_features = 10
n_tasks = 17
tasks = range(n_tasks)
# Define train dataset
n_train = 100
X_train = np.random.rand(n_train, n_features)
y_train = np.random.randint(2, size=(n_train, n_tasks))
w_train = np.ones_like(y_train)
ids_train = ["C"] * n_train
train_dataset = DiskDataset.from_numpy(self.train_dir, X_train,
y_train, w_train, ids_train)
# Define test dataset
n_test = 10
X_test = np.random.rand(n_test, n_features)
y_test = np.random.randint(2, size=(n_test, n_tasks))
w_test = np.ones_like(y_test)
ids_test = ["C"] * n_test
test_dataset = DiskDataset.from_numpy(self.test_dir, X_test, y_test,
w_test, ids_test)
transformers = []
classification_metrics = [Metric(metrics.roc_auc_score)]
def model_builder(model_dir):
sklearn_model = LogisticRegression()
return SklearnModel(sklearn_model, model_dir)
multitask_model = SingletaskToMultitask(tasks, model_builder,
self.model_dir)
# Fit trained model
multitask_model.fit(train_dataset)
multitask_model.save()
# Eval multitask_model on train
evaluator = Evaluator(multitask_model,
train_dataset,
transformers,
verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
# Eval multitask_model on test
evaluator = Evaluator(multitask_model,
test_dataset,
transformers,
verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
def test_sklearn_transformed_regression(self):
"""Test that sklearn models can learn on simple transformed regression datasets."""
np.random.seed(123)
dataset = sklearn.datasets.load_diabetes()
X, y = dataset.data, dataset.target
frac_train = .7
n_samples = len(X)
n_train = int(frac_train * n_samples)
X_train, y_train = X[:n_train], y[:n_train]
X_test, y_test = X[n_train:], y[n_train:]
train_dataset = DiskDataset.from_numpy(self.train_dir, X_train,
y_train)
test_dataset = DiskDataset.from_numpy(self.test_dir, X_test, y_test)
# Eval model on train
transformers = [
NormalizationTransformer(transform_X=True, dataset=train_dataset),
ClippingTransformer(transform_X=True, dataset=train_dataset),
NormalizationTransformer(transform_y=True, dataset=train_dataset)
]
for data in [train_dataset, test_dataset]:
for transformer in transformers:
transformer.transform(data)
verbosity = "high"
regression_metric = Metric(metrics.r2_score, verbosity=verbosity)
sklearn_model = LinearRegression()
model = SklearnModel(sklearn_model, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
train_evaluator = Evaluator(model,
train_dataset,
transformers,
verbosity=verbosity)
train_scores = train_evaluator.compute_model_performance(
[regression_metric])
assert train_scores[regression_metric.name] > .5
# Eval model on test
evaluator = Evaluator(model,
test_dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] > .5
def test_tf_skewed_classification_overfit(self):
"""Test tensorflow models can overfit 0/1 datasets with few actives."""
tasks = ["task0"]
task_types = {task: "classification" for task in tasks}
#n_samples = 100
n_samples = 100
n_features = 3
n_tasks = len(tasks)
n_classes = 2
# Generate dummy dataset
np.random.seed(123)
p = .05
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.binomial(1, p, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"layer_sizes": [1500],
"dropouts": [.0],
"learning_rate": 0.003,
"momentum": .9,
"batch_size": n_samples,
"num_classification_tasks": 1,
"num_classes": n_classes,
"num_features": n_features,
"weight_init_stddevs": [1.],
"bias_init_consts": [1.],
"nb_epoch": 200,
"penalty": 0.0,
"optimizer": "adam",
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
model = TensorflowModel(tasks,
task_types,
model_params,
self.model_dir,
tf_class=TensorflowMultiTaskClassifier,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .8
def test_tf_multitask_regression_overfit(self):
"""Test tf multitask overfits tiny data."""
n_tasks = 10
n_samples = 10
n_features = 3
n_classes = 2
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.zeros((n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
regression_metric = Metric(metrics.mean_squared_error, verbosity=verbosity,
task_averager=np.mean, mode="regression")
tensorflow_model = TensorflowMultiTaskRegressor(
n_tasks, n_features, self.model_dir, dropouts=[0.],
learning_rate=0.0003, weight_init_stddevs=[.1],
batch_size=n_samples, verbosity=verbosity)
model = TensorflowModel(tensorflow_model, self.model_dir)
# Fit trained model
model.fit(dataset, nb_epoch=50)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] < .1
def evaluate(self,
dataset,
metrics,
transformers=[],
per_task_metrics=False):
"""
Evaluates the performance of this model on specified dataset.
Parameters
----------
dataset: dc.data.Dataset
Dataset object.
metric: deepchem.metrics.Metric
Evaluation metric
transformers: list
List of deepchem.transformers.Transformer
per_task_metrics: bool
If True, return per-task scores.
Returns
-------
dict
Maps tasks to scores under metric.
"""
evaluator = Evaluator(self, dataset, transformers)
if not per_task_metrics:
scores = evaluator.compute_model_performance(metrics)
return scores
else:
scores, per_task_scores = evaluator.compute_model_performance(
metrics, per_task_metrics=per_task_metrics)
return scores, per_task_scores
def test_sklearn_multitask_regression_overfit(self):
"""Test SKLearn singletask-to-multitask overfits tiny regression data."""
n_tasks = 2
tasks = ["task%d" % task for task in range(n_tasks)]
n_samples = 10
n_features = 3
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.rand(n_samples, n_tasks)
w = np.ones((n_samples, n_tasks))
dataset = DiskDataset.from_numpy(self.train_dir, X, y, w, ids)
verbosity = "high"
regression_metric = Metric(metrics.r2_score, verbosity=verbosity, task_averager=np.mean)
def model_builder(model_dir):
sklearn_model = RandomForestRegressor()
return SklearnModel(sklearn_model, model_dir)
model = SingletaskToMultitask(tasks, model_builder, self.model_dir)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] > .7
def test_keras_multitask_regression_overfit(self):
"""Test keras multitask overfits tiny data."""
g = tf.Graph()
sess = tf.Session(graph=g)
K.set_session(sess)
with g.as_default():
n_tasks = 10
n_samples = 10
n_features = 3
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
regression_metric = Metric(metrics.r2_score, verbosity=verbosity,
task_averager=np.mean, mode="regression")
keras_model = MultiTaskDNN(n_tasks, n_features, "regression",
dropout=0., learning_rate=.1, decay=1e-4)
model = KerasModel(keras_model, self.model_dir, verbosity=verbosity)
# Fit trained model
model.fit(dataset, nb_epoch=100)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] > .75
def test_tf_regression_overfit(self):
"""Test that TensorFlow models can overfit simple regression datasets."""
n_samples = 10
n_features = 3
n_tasks = 1
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.zeros((n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
regression_metric = Metric(metrics.mean_squared_error, verbosity=verbosity)
# TODO(rbharath): This breaks with optimizer="momentum". Why?
tensorflow_model = TensorflowMultiTaskRegressor(
n_tasks, n_features, self.model_dir, dropouts=[0.],
learning_rate=0.003, weight_init_stddevs=[np.sqrt(6)/np.sqrt(1000)],
batch_size=n_samples, verbosity=verbosity)
model = TensorflowModel(tensorflow_model, self.model_dir)
# Fit trained model
model.fit(dataset, nb_epoch=100)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] < .1
def test_tf_classification_overfit(self):
"""Test that tensorflow models can overfit simple classification datasets."""
n_samples = 10
n_features = 3
n_tasks = 1
n_classes = 2
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.zeros((n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
classification_metric = Metric(metrics.accuracy_score, verbosity=verbosity)
tensorflow_model = TensorflowMultiTaskClassifier(
n_tasks, n_features, self.model_dir, dropouts=[0.],
learning_rate=0.0003, weight_init_stddevs=[.1],
batch_size=n_samples, verbosity=verbosity)
model = TensorflowModel(tensorflow_model, self.model_dir)
# Fit trained model
model.fit(dataset, nb_epoch=100)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_sklearn_classification_overfit(self):
"""Test that sklearn models can overfit simple classification datasets."""
n_samples = 10
n_features = 3
n_tasks = 1
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score, verbosity=verbosity)
sklearn_model = RandomForestClassifier()
model = SklearnModel(sklearn_model, self.model_dir)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_singletask_sklearn_rf_RDKIT_descriptor_regression_API(self):
"""Test of singletask RF RDKIT-descriptor regression API."""
splittype = "scaffold"
featurizer = RDKitDescriptors()
tasks = ["log-solubility"]
task_type = "regression"
task_types = {task: task_type for task in tasks}
input_file = os.path.join(self.current_dir, "example.csv")
loader = DataLoader(tasks=tasks,
smiles_field=self.smiles_field,
featurizer=featurizer,
verbosity="low")
dataset = loader.featurize(input_file, self.data_dir)
splitter = ScaffoldSplitter()
train_dataset, test_dataset = splitter.train_test_split(
dataset, self.train_dir, self.test_dir)
input_transformers = [
NormalizationTransformer(transform_X=True, dataset=train_dataset),
ClippingTransformer(transform_X=True, dataset=train_dataset)]
output_transformers = [
NormalizationTransformer(transform_y=True, dataset=train_dataset)]
transformers = input_transformers + output_transformers
for dataset in [train_dataset, test_dataset]:
for transformer in transformers:
transformer.transform(dataset)
regression_metrics = [Metric(metrics.r2_score),
Metric(metrics.mean_squared_error),
Metric(metrics.mean_absolute_error)]
sklearn_model = RandomForestRegressor()
model = SklearnModel(sklearn_model, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
evaluator = Evaluator(model, train_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(regression_metrics)
# Eval model on test
evaluator = Evaluator(model, test_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(regression_metrics)
def test_sklearn_multitask_classification_overfit(self):
"""Test SKLearn singletask-to-multitask overfits tiny data."""
n_tasks = 10
tasks = ["task%d" % task for task in range(n_tasks)]
task_types = {task: "classification" for task in tasks}
n_samples = 10
n_features = 3
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"batch_size": None,
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
def model_builder(tasks,
task_types,
model_params,
model_dir,
verbosity=None):
return SklearnModel(tasks,
task_types,
model_params,
model_dir,
mode="classification",
model_instance=RandomForestClassifier(),
verbosity=verbosity)
model = SingletaskToMultitask(tasks,
task_types,
model_params,
self.model_dir,
model_builder,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_graph_conv_singletask_classification_overfit(self):
"""Test graph-conv multitask overfits tiny data."""
g = tf.Graph()
sess = tf.Session(graph=g)
K.set_session(sess)
with g.as_default():
n_tasks = 1
n_samples = 10
n_features = 3
n_classes = 2
# Load mini log-solubility dataset.
splittype = "scaffold"
featurizer = ConvMolFeaturizer()
tasks = ["outcome"]
task_type = "classification"
task_types = {task: task_type for task in tasks}
input_file = os.path.join(self.current_dir, "example_classification.csv")
loader = DataLoader(tasks=tasks,
smiles_field=self.smiles_field,
featurizer=featurizer,
verbosity="low")
dataset = loader.featurize(input_file, self.data_dir)
verbosity = "high"
classification_metric = Metric(metrics.accuracy_score, verbosity=verbosity)
#n_atoms = 50
n_feat = 71
batch_size = 10
graph_model = SequentialGraphModel(n_feat)
graph_model.add(GraphConv(64, activation='relu'))
graph_model.add(BatchNormalization(epsilon=1e-5, mode=1))
graph_model.add(GraphPool())
# Gather Projection
graph_model.add(Dense(128, activation='relu'))
graph_model.add(BatchNormalization(epsilon=1e-5, mode=1))
graph_model.add(GraphGather(batch_size, activation="tanh"))
with self.test_session() as sess:
model = MultitaskGraphClassifier(
sess, graph_model, n_tasks, self.model_dir, batch_size=batch_size,
learning_rate=1e-3, learning_rate_decay_time=1000,
optimizer_type="adam", beta1=.9, beta2=.999, verbosity="high")
# Fit trained model
model.fit(dataset, nb_epoch=20)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
######################################################### DEBUG
print("scores")
print(scores)
######################################################### DEBUG
assert scores[classification_metric.name] > .85
def test_tf_multitask_classification_overfit(self):
"""Test tf multitask overfits tiny data."""
n_tasks = 10
tasks = ["task%d" % task for task in range(n_tasks)]
task_types = {task: "classification" for task in tasks}
n_samples = 10
n_features = 3
n_classes = 2
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
#y = np.random.randint(n_classes, size=(n_samples, n_tasks))
y = np.zeros((n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"layer_sizes": [1000],
"dropouts": [.0],
"learning_rate": 0.0003,
"momentum": .9,
"batch_size": n_samples,
"num_classification_tasks": n_tasks,
"num_classes": n_classes,
"num_features": n_features,
"weight_init_stddevs": [.1],
"bias_init_consts": [1.],
"nb_epoch": 100,
"penalty": 0.0,
"optimizer": "adam",
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
classification_metric = Metric(metrics.accuracy_score,
verbosity=verbosity)
model = TensorflowModel(tasks,
task_types,
model_params,
self.model_dir,
tf_class=TensorflowMultiTaskClassifier,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_singletask_tf_mlp_ECFP_classification_API(self):
"""Straightforward test of Tensorflow singletask deepchem classification API."""
n_features = 1024
featurizer = CircularFingerprint(size=n_features)
tasks = ["outcome"]
input_file = os.path.join(self.current_dir, "example_classification.csv")
loader = DataLoader(tasks=tasks,
smiles_field=self.smiles_field,
featurizer=featurizer,
verbosity="low")
dataset = loader.featurize(input_file, self.data_dir)
splitter = ScaffoldSplitter()
train_dataset, test_dataset = splitter.train_test_split(
dataset, self.train_dir, self.test_dir)
transformers = [
NormalizationTransformer(transform_y=True, dataset=train_dataset)]
for dataset in [train_dataset, test_dataset]:
for transformer in transformers:
transformer.transform(dataset)
classification_metrics = [Metric(metrics.roc_auc_score),
Metric(metrics.matthews_corrcoef),
Metric(metrics.recall_score),
Metric(metrics.accuracy_score)]
tensorflow_model = TensorflowMultiTaskClassifier(
len(tasks), n_features, self.model_dir)
model = TensorflowModel(tensorflow_model, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
evaluator = Evaluator(model, train_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
# Eval model on test
evaluator = Evaluator(model, test_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
def _evaluate(self,model,metrics,output_transformers,*datasets):
scores=[]
for dataset in datasets:
evaluator= Evaluator(
model, dataset, output_transformers)
score = evaluator.compute_model_performance(metrics)
scores.append(score)
# print(scores)
return scores
def test_multitask_keras_mlp_ECFP_classification_API(self):
"""Straightforward test of Keras multitask deepchem classification API."""
g = tf.Graph()
sess = tf.Session(graph=g)
K.set_session(sess)
with g.as_default():
task_type = "classification"
input_file = os.path.join(self.current_dir, "multitask_example.csv")
tasks = ["task0", "task1", "task2", "task3", "task4", "task5", "task6",
"task7", "task8", "task9", "task10", "task11", "task12",
"task13", "task14", "task15", "task16"]
n_features = 1024
featurizer = CircularFingerprint(size=n_features)
loader = DataLoader(tasks=tasks,
smiles_field=self.smiles_field,
featurizer=featurizer,
verbosity="low")
dataset = loader.featurize(input_file, self.data_dir)
splitter = ScaffoldSplitter()
train_dataset, test_dataset = splitter.train_test_split(
dataset, self.train_dir, self.test_dir)
transformers = []
classification_metrics = [Metric(metrics.roc_auc_score),
Metric(metrics.matthews_corrcoef),
Metric(metrics.recall_score),
Metric(metrics.accuracy_score)]
keras_model = MultiTaskDNN(len(tasks), n_features, "classification",
dropout=0.)
model = KerasModel(keras_model, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
evaluator = Evaluator(model, train_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
# Eval model on test
evaluator = Evaluator(model, test_dataset, transformers, verbosity=True)
_ = evaluator.compute_model_performance(classification_metrics)
def test_keras_skewed_classification_overfit(self):
"""Test keras models can overfit 0/1 datasets with few actives."""
tasks = ["task0"]
task_types = {task: "classification" for task in tasks}
n_samples = 100
n_features = 3
n_tasks = len(tasks)
# Generate dummy dataset
np.random.seed(123)
p = .05
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.binomial(1, p, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"nb_hidden": 1000,
"activation": "relu",
"dropout": .0,
"learning_rate": .15,
"momentum": .9,
"nesterov": False,
"decay": 1e-4,
"batch_size": n_samples,
"nb_epoch": 200,
"init": "glorot_uniform",
"nb_layers": 1,
"batchnorm": False,
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
model = MultiTaskDNN(tasks,
task_types,
model_params,
self.model_dir,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_tf_regression_overfit(self):
"""Test that TensorFlow models can overfit simple regression datasets."""
tasks = ["task0"]
task_types = {task: "regression" for task in tasks}
n_samples = 10
n_features = 3
n_tasks = len(tasks)
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.zeros((n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"layer_sizes": [1000],
"dropouts": [.0],
"learning_rate": 0.003,
"momentum": .9,
"batch_size": n_samples,
"num_regression_tasks": 1,
"num_features": n_features,
"weight_init_stddevs": [np.sqrt(6) / np.sqrt(1000)],
"bias_init_consts": [1.],
"nb_epoch": 100,
"penalty": 0.0,
"optimizer": "momentum",
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
regression_metric = Metric(metrics.mean_squared_error,
verbosity=verbosity)
model = TensorflowModel(tasks,
task_types,
model_params,
self.model_dir,
tf_class=TensorflowMultiTaskRegressor,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] < .1
def test_sklearn_classification(self):
"""Test that sklearn models can learn on simple classification datasets."""
np.random.seed(123)
dataset = sklearn.datasets.load_digits(n_class=2)
X, y = dataset.data, dataset.target
frac_train = .7
n_samples = len(X)
n_train = int(frac_train * n_samples)
X_train, y_train = X[:n_train], y[:n_train]
X_test, y_test = X[n_train:], y[n_train:]
train_dataset = DiskDataset.from_numpy(self.train_dir, X_train,
y_train)
test_dataset = DiskDataset.from_numpy(self.test_dir, X_test, y_test)
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
sklearn_model = LogisticRegression()
model = SklearnModel(sklearn_model, self.model_dir)
# Fit trained model
model.fit(train_dataset)
model.save()
# Eval model on train
transformers = []
train_evaluator = Evaluator(model,
train_dataset,
transformers,
verbosity=verbosity)
train_scores = train_evaluator.compute_model_performance(
[classification_metric])
# Eval model on test
transformers = []
evaluator = Evaluator(model,
test_dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .5
def test_keras_multitask_regression_overfit(self):
"""Test keras multitask overfits tiny data."""
n_tasks = 10
tasks = ["task%d" % task for task in range(n_tasks)]
task_types = {task: "regression" for task in tasks}
n_samples = 10
n_features = 3
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"nb_hidden": 1000,
"activation": "relu",
"dropout": .0,
"learning_rate": .15,
"momentum": .9,
"nesterov": False,
"decay": 1e-4,
"batch_size": n_samples,
"nb_epoch": 200,
"init": "glorot_uniform",
"nb_layers": 1,
"batchnorm": False,
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
regression_metric = Metric(metrics.r2_score, verbosity=verbosity)
model = MultiTaskDNN(tasks,
task_types,
model_params,
self.model_dir,
verbosity=verbosity)
# Fit trained model
model.fit(dataset)
model.save()
# Eval model on train
transformers = []
evaluator = Evaluator(model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([regression_metric])
assert scores[regression_metric.name] > .9
def test_multiclass_classification_singletask():
"""Test multiclass classification evaluation."""
X = np.random.rand(100, 5)
y = np.random.randint(5, size=(100,))
dataset = dc.data.NumpyDataset(X, y)
model = dc.models.MultitaskClassifier(1, 5, n_classes=5)
evaluator = Evaluator(model, dataset, [])
multitask_scores = evaluator.compute_model_performance(
dc.metrics.roc_auc_score, n_classes=5)
assert len(multitask_scores) == 1
assert multitask_scores["metric-1"] >= 0
def test_sklearn_reload(self):
"""Test that trained model can be reloaded correctly."""
tasks = ["task0"]
task_types = {task: "classification" for task in tasks}
n_samples = 10
n_features = 3
n_tasks = len(tasks)
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = Dataset.from_numpy(self.train_dir, X, y, w, ids, tasks)
model_params = {
"batch_size": None,
"data_shape": dataset.get_data_shape()
}
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
model = SklearnModel(tasks,
task_types,
model_params,
self.model_dir,
mode="classification",
model_instance=RandomForestClassifier())
# Fit trained model
model.fit(dataset)
model.save()
# Load trained model
reloaded_model = SklearnModel(tasks,
task_types,
model_params,
self.model_dir,
mode="classification")
reloaded_model.reload()
# Eval model on train
transformers = []
evaluator = Evaluator(reloaded_model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance([classification_metric])
assert scores[classification_metric.name] > .9
def test_keras_reload(self):
"""Test that trained keras models can be reloaded correctly."""
g = tf.Graph()
sess = tf.Session(graph=g)
K.set_session(sess)
with g.as_default():
tasks = ["task0"]
task_types = {task: "classification" for task in tasks}
n_samples = 10
n_features = 3
n_tasks = len(tasks)
# Generate dummy dataset
np.random.seed(123)
ids = np.arange(n_samples)
X = np.random.rand(n_samples, n_features)
y = np.random.randint(2, size=(n_samples, n_tasks))
w = np.ones((n_samples, n_tasks))
dataset = NumpyDataset(X, y, w, ids)
verbosity = "high"
classification_metric = Metric(metrics.roc_auc_score,
verbosity=verbosity)
keras_model = MultiTaskDNN(n_tasks,
n_features,
"classification",
dropout=0.)
model = KerasModel(keras_model, self.model_dir)
# Fit trained model
model.fit(dataset)
model.save()
# Load trained model
reloaded_keras_model = MultiTaskDNN(n_tasks,
n_features,
"classification",
dropout=0.)
reloaded_model = KerasModel(reloaded_keras_model, self.model_dir)
reloaded_model.reload(
custom_objects={"MultiTaskDNN": MultiTaskDNN})
# Eval model on train
transformers = []
evaluator = Evaluator(reloaded_model,
dataset,
transformers,
verbosity=verbosity)
scores = evaluator.compute_model_performance(
[classification_metric])
assert scores[classification_metric.name] > .6
def test_evaluator_dc_metric(): """Test an evaluator on a dataset.""" X = np.random.rand(10, 5) y = np.random.rand(10, 1) dataset = dc.data.NumpyDataset(X, y) model = dc.models.MultitaskRegressor(1, 5) evaluator = Evaluator(model, dataset, []) metric = dc.metrics.Metric(dc.metrics.mae_score) multitask_scores = evaluator.compute_model_performance(metric) assert isinstance(multitask_scores, dict) assert len(multitask_scores) == 1 assert multitask_scores['mae_score'] > 0
def _create_model(self, train_dataset, test_dataset, model):
"""Helper method to create model for test."""
# Fit model
model.fit(train_dataset)
model.save(self.model_dir)
# Eval model on train
evaluator = Evaluator(model, train_dataset, verbose=True)
with tempfile.NamedTemporaryFile() as train_csv_out:
with tempfile.NamedTemporaryFile() as train_stats_out:
_, _ = evaluator.compute_model_performance(
train_csv_out, train_stats_out)
# Eval model on test
evaluator = Evaluator(model, test_dataset, verbose=True)
with tempfile.NamedTemporaryFile() as test_csv_out:
with tempfile.NamedTemporaryFile() as test_stats_out:
_, _ = evaluator.compute_model_performance(
test_csv_out, test_stats_out)
