def testDQNCartPole(self):
root_dir = self.get_temp_dir()
train_loss = train_eval.train_eval(root_dir,
num_iterations=1,
num_eval_episodes=1,
initial_collect_steps=10)
self.assertGreater(train_loss.loss, 0.0)
def testRNNDQNMaskedCartPole(self):
root_dir = self.get_temp_dir()
train_loss = train_eval.train_eval(root_dir,
env_name='MaskedCartPole-v0',
train_sequence_length=2,
initial_collect_steps=10,
num_eval_episodes=1,
num_iterations=1)
self.assertGreater(train_loss.loss, 0.0)
def testDQNCartPole(self):
if not tf.executing_eagerly():
self.skipTest('Binary is eager-only.')
root_dir = self.get_temp_dir()
train_loss = train_eval.train_eval(root_dir,
num_iterations=1,
num_eval_episodes=1,
initial_collect_steps=10)
self.assertGreater(train_loss.loss, 0.0)
def testRNNDQNMaskedCartPole(self):
if not tf.executing_eagerly():
self.skipTest('Binary is eager-only.')
root_dir = self.get_temp_dir()
train_loss = train_eval.train_eval(root_dir,
env_name='MaskedCartPole-v0',
train_sequence_length=2,
initial_collect_steps=10,
num_eval_episodes=1,
num_iterations=1)
self.assertGreater(train_loss.loss, 0.0)