def testSynchronousTrainCollectEval(self):
"""End-to-end integration test.
"""
env = grasping_env.KukaGraspingProceduralEnv(downsample_width=64,
downsample_height=64,
continuous=True,
remove_height_hack=True,
render_mode='DIRECT')
data_dir = 'testdata'
gin_config = os.path.join(FLAGS.test_srcdir, data_dir, 'random_collect.gin')
# Collect initial data from random policy without training.
with open(gin_config, 'r') as f:
gin.parse_config(f)
train_collect_eval.train_collect_eval(collect_env=env,
eval_env=None,
test_env=None,
root_dir=self._root_dir,
train_fn=None)
# Run training (synchronous train, collect, & eval).
gin_config = os.path.join(FLAGS.test_srcdir, data_dir, 'train_dqn.gin')
with open(gin_config, 'r') as f:
gin.parse_config(f)
train_collect_eval.train_collect_eval(collect_env=env,
eval_env=None,
test_env=None,
root_dir=self._root_dir)
def resetEnvironment():
global obs, env
env = grasping_env.KukaGraspingProceduralEnv(downsample_width=48,
downsample_height=48,
continuous=True,
remove_height_hack=True,
render_mode='GUI')
print(env.action_space)
obs = env.reset()
done, env_step, episode_reward, episode_data = (False, 0, 0.0, [])
def testDDPGPolicy(self):
np.random.seed(0)
env = grasping_env.KukaGraspingProceduralEnv(
downsample_width=48, downsample_height=48,
continuous=True, remove_height_hack=True, render_mode='DIRECT')
policy = policies.DDPGPolicy(tf_critics.cnn_v0,
tf_critics.cnn_ia_v1,
state_shape=(1, 48, 48, 3),
action_size=4,
use_gpu=False,
build_target=False,
include_timestep=True)
policy.reset()
obs = env.reset()
action, debug = policy.sample_action(obs, 0)
self.assertLen(action, 4)
self.assertIn('q', debug)
def testPolicyRun(self, tag, use_root_dir):
env = grasping_env.KukaGraspingProceduralEnv(downsample_width=48,
downsample_height=48,
continuous=True,
remove_height_hack=True,
render_mode='DIRECT')
policy = policies.RandomGraspingPolicyD4()
root_dir = os.path.join(FLAGS.test_tmpdir,
tag) if use_root_dir else None
run_env.run_env(env,
policy=policy,
explore_schedule=None,
episode_to_transitions_fn=None,
replay_writer=None,
root_dir=root_dir,
tag=tag,
task=0,
global_step=0,
num_episodes=1)
