Adding a unit test (b6beb3bf) · Commits · 钟慕尧 / deepchem

deepchem/models/keras_model.py

+29 −14

Original line number	Diff line number	Diff line
		@@ -272,7 +272,8 @@ class KerasModel(Model):
		restore: bool = False,
		variables: Optional[List[tf.Variable]] = None,
		loss: Optional[KerasLossFn] = None,
		callbacks: Union[Callable, List[Callable]] = []) -> List[float]:
		callbacks: Union[Callable, List[Callable]] = [],
		return_loss_curve: bool = False) -> Union[float, List[float]]:
		"""Train this model on a dataset.

		Parameters
		@@ -302,16 +303,20 @@ class KerasModel(Model):
		callbacks: function or list of functions
		one or more functions of the form f(model, step) that will be invoked after
		every step. This can be used to perform validation, logging, etc.
		return_loss_curve: bool, optional (default False)
		If `True` return the full set of average losses computed over the
		process of fitting. Else return the last computed average loss.

		Returns
		-------
		the average loss over the most recent checkpoint interval
		Either the average loss over the most recent checkpoint interval or a list
		of all such average losses over the course of fitting.
		"""
		return self.fit_generator(
		self.default_generator(
		dataset, epochs=nb_epoch,
		deterministic=deterministic), max_checkpoints_to_keep,
		checkpoint_interval, restore, variables, loss, callbacks)
		dataset, epochs=nb_epoch, deterministic=deterministic),
		max_checkpoints_to_keep, checkpoint_interval, restore, variables, loss,
		callbacks, return_loss_curve)

		def fit_generator(
		self,
		@@ -321,7 +326,8 @@ class KerasModel(Model):
		restore: bool = False,
		variables: Optional[List[tf.Variable]] = None,
		loss: Optional[KerasLossFn] = None,
		callbacks: Union[Callable, List[Callable]] = []) -> List[float]:
		callbacks: Union[Callable, List[Callable]] = [],
		return_loss_curve: bool = False) -> Union[float, List[float]]:
		"""Train this model on data from a generator.

		Parameters
		@@ -347,10 +353,14 @@ class KerasModel(Model):
		callbacks: function or list of functions
		one or more functions of the form f(model, step) that will be invoked after
		every step. This can be used to perform validation, logging, etc.
		return_loss_curve: bool, optional (default False)
		If `True` return the full set of average losses computed over the
		process of fitting. Else return the last computed average loss.

		Returns
		-------
		the average loss over the most recent checkpoint interval
		Either the average loss over the most recent checkpoint interval or a list
		of all such average losses over the course of fitting.
		"""
		if not isinstance(callbacks, SequenceCollection):
		callbacks = [callbacks]
		@@ -431,7 +441,13 @@ class KerasModel(Model):

		time2 = time.time()
		logger.info("TIMING: model fitting took %0.3f s" % (time2 - time1))
		if return_loss_curve:
		return avg_losses
		else:
		if len(avg_losses) > 0:
		return avg_losses[-1]
		else:
		return 0.0

		def _create_gradient_fn(self,
		variables: Optional[List[tf.Variable]]) -> Callable:
		@@ -500,18 +516,17 @@ class KerasModel(Model):
		"""
		self._ensure_built()
		dataset = NumpyDataset(X, y, w)
		losses = self.fit(
		# We set return_loss_curve=False, so we know this is a float, but mypy
		# can't automatically infer that.
		return self.fit( # type: ignore
		dataset,
		nb_epoch=1,
		max_checkpoints_to_keep=max_checkpoints_to_keep,
		checkpoint_interval=self._global_step.numpy() + 2 if checkpoint else 0,
		variables=variables,
		loss=loss,
		callbacks=callbacks)
		if len(losses) != 1:
		raise ValueError(
		"Each batch should take only one global step to fit. Unknown error.")
		return losses[0]
		callbacks=callbacks,
		return_loss_curve=False)

		def _predict(
		self, generator: Iterable[Tuple[Any, Any, Any]],

deepchem/models/models.py

+3 −2

Original line number	Diff line number	Diff line
		@@ -21,7 +21,7 @@ from deepchem.utils.save import load_from_disk
		from deepchem.utils.save import save_to_disk
		from deepchem.utils.evaluate import Evaluator

		from typing import Any, Dict, List, Optional, Sequence
		from typing import Any, Dict, List, Optional, Sequence, Union
		from deepchem.utils.typing import OneOrMany

		logger = logging.getLogger(__name__)
		@@ -127,7 +127,8 @@ class Model(BaseEstimator):
		"""
		raise NotImplementedError

		def fit(self, dataset: Dataset, nb_epoch: int = 10) -> List[float]:
		def fit(self, dataset: Dataset,
		nb_epoch: int = 10) -> Union[float, List[float]]:
		"""
		Fits a model on data in a Dataset object.

deepchem/models/tests/test_kerasmodel.py

+21 −0

Original line number	Diff line number	Diff line
		@@ -58,6 +58,27 @@ def test_overfit_sequential_model():
		assert scores[metric.name] > 0.9


		def test_fit_return_loss_curve():
		"""Test fitting a KerasModel and getting a loss curve back."""
		n_data_points = 10
		n_features = 2
		X = np.random.rand(n_data_points, n_features)
		y = (X[:, 0] > X[:, 1]).astype(np.float32)
		dataset = dc.data.NumpyDataset(X, y)
		keras_model = tf.keras.Sequential([
		tf.keras.layers.Dense(10, activation='relu'),
		tf.keras.layers.Dense(1, activation='sigmoid')
		])
		model = dc.models.KerasModel(
		keras_model,
		dc.models.losses.BinaryCrossEntropy(),
		learning_rate=0.005,
		log_frequency=10)
		losses = model.fit(dataset, nb_epoch=1000, return_loss_curve=True)
		# Each epoch is a single step for this model
		assert len(losses) == 100


		def test_fit_on_batch():
		"""Test fitting a KerasModel to individual batches."""
		n_data_points = 10

Admin message