first (7eca4ad6) · Commits · 钟慕尧 / deepchem

deepchem/rl/init.py

+14 −2

Original line number	Diff line number	Diff line
		"""Interface for reinforcement learning."""

		from deepchem.utils import TensorFlowStub

		try:
		import tensorflow as tf
		import tensorflow_probability as tfp
		from deepchem.rl.a2c import A2C
		from deepchem.rl.ppo import PPO
		except ModuleNotFoundError:

		class A2C(TensorFlowStub):
		pass

		class PPO(TensorFlowStub):
		pass


		class Environment(object):

deepchem/rl/a2c.py

+24 −7

Original line number	Diff line number	Diff line
		"""Advantage Actor-Critic (A2C) algorithm for reinforcement learning."""

		from deepchem.models import KerasModel
		from deepchem.models.optimizers import Adam
		import numpy as np
		import tensorflow as tf
		import collections
		import copy
		import multiprocessing
		@@ -11,19 +8,32 @@ import os
		import re
		import threading
		import time
		from deepchem.models import KerasModel
		from deepchem.models.optimizers import Adam


		class A2CLossDiscrete(object):
		"""This class computes the loss function for A2C with discrete action spaces."""
		"""This class computes the loss function for A2C with discrete action spaces.

		Note
		----
		This class requires tensorflow to be installed.
		"""

		def __init__(self, value_weight, entropy_weight, action_prob_index,
		value_index):

		try:
		import tensorflow as tf
		except ModuleNotFoundError:
		raise ValueError("This class requires tensorflow to be installed.")
		self.value_weight = value_weight
		self.entropy_weight = entropy_weight
		self.action_prob_index = action_prob_index
		self.value_index = value_index

		def __call__(self, outputs, labels, weights):
		import tensorflow as tf
		prob = outputs[self.action_prob_index]
		value = outputs[self.value_index]
		reward, advantage = weights
		@@ -43,16 +53,18 @@ class A2CLossContinuous(object):

		Note
		----
		This class requires tensorflow-probability to be installed.
		This class requires tensorflow, tensorflow-probability to be installed.
		"""

		def __init__(self, value_weight, entropy_weight, mean_index, std_index,
		value_index):
		try:
		import tensorflow as tf
		import tensorflow_probability as tfp
		except ModuleNotFoundError:
		raise ValueError(
		"This class requires tensorflow-probability to be installed.")
		"This class requires tensorflow, tensorflow-probability to be installed."
		)
		self.value_weight = value_weight
		self.entropy_weight = entropy_weight
		self.mean_index = mean_index
		@@ -126,7 +138,8 @@ class A2C(object):

		Note
		----
		Using this class on continuous action spaces requires that `tensorflow_probability` be installed.
		This class requires tensorflow to be installed. Using this class on
		continuous action spaces requires that `tensorflow_probability` be installed.
		"""

		def __init__(self,
		@@ -166,6 +179,10 @@ class A2C(object):
		the directory in which the model will be saved. If None, a temporary directory will be created.
		use_hindsight: bool
		if True, use Hindsight Experience Replay

		Note
		----
		This class requires tensorflow to be installed.
		"""
		self._env = env
		self._policy = policy

deepchem/rl/ppo.py

+16 −2

Original line number	Diff line number	Diff line
		@@ -14,7 +14,12 @@ import time


		class PPOLoss(object):
		"""This class computes the loss function for PPO."""
		"""This class computes the loss function for PPO.

		Note
		----
		This class requires tensorflow to be installed.
		"""

		def __init__(self, value_weight, entropy_weight, clipping_width,
		action_prob_index, value_index):
		@@ -138,6 +143,10 @@ class PPO(object):
		the directory in which the model will be saved. If None, a temporary directory will be created.
		use_hindsight: bool
		if True, use Hindsight Experience Replay

		Note
		----
		This class requires tensorflow to be installed.
		"""
		self._env = env
		self._policy = policy
		@@ -405,7 +414,12 @@ class PPO(object):


		class _Worker(object):
		"""A Worker object is created for each training thread."""
		"""A Worker object is created for each training thread.

		Note
		----
		This class requires tensorflow to be installed.
		"""

		def __init__(self, ppo, index):
		self.ppo = ppo

deepchem/utils/init.py

+9 −0

Original line number	Diff line number	Diff line
		@@ -21,6 +21,15 @@ except:
		from urllib import urlretrieve # Python 2


		class TensorFlowStub(object):
		"""This class provides a stub to handle TensorFlow import failures."""

		def __init__(self, args, *kwargs):
		raise ModuleNotFoundError(
		"The class '%s' cannot be used because TensorFlow is not installed" %
		type(self).__name__)


		def pad_array(x, shape, fill=0, both=False):
		"""
		Pad an array with a fill value.

Admin message