def test_train_agent_dqn(self):
config = TestAgentConfig.get()
experiment = Experiment(config)
experiment.set_env('lunarlander-dqn')
scores = experiment.train()
assert len(scores) == (
config.get_current_exp_cfg().trainer_cfg.max_steps /
config.get_current_exp_cfg().trainer_cfg.max_episode_steps)
def test_train_agent_dqn_with_disabled_prioritized_replay(self):
config = TestAgentConfig.get()
experiment = Experiment(config)
experiment.set_env('lunarlander-dqn-noprio')
scores = experiment.train()
assert len(scores) == (
config.get_current_exp_cfg().trainer_cfg.max_steps /
config.get_current_exp_cfg().trainer_cfg.max_episode_steps)
def test_train_environment_gymai_lunarlander(self):
config = TestEnvConfig.get()
experiment = Experiment(config)
experiment.set_env('lunarlander')
scores = experiment.train()
assert len(scores) == (
config.get_current_exp_cfg().trainer_cfg.max_steps /
config.get_current_exp_cfg().trainer_cfg.max_episode_steps)
def test_train_environment_unity_banana(self):
config = TestEnvConfig.get()
experiment = Experiment(config)
experiment.set_env('banana')
scores = experiment.train()
assert len(scores) == (
config.get_current_exp_cfg().trainer_cfg.max_steps /
config.get_current_exp_cfg().trainer_cfg.max_episode_steps)
def test_train_environment_atari_breakout_rgb(self):
config = TestEnvConfig.get()
experiment = Experiment(config)
experiment.set_env('breakout-rgb')
scores = experiment.train()
assert len(scores) == (
config.get_current_exp_cfg().trainer_cfg.max_steps /
config.get_current_exp_cfg().trainer_cfg.max_episode_steps)
def test_train_configExist_canTrain1Episode(self):
config = Config(test=True)
experiment = Experiment(config)
envs = experiment.list_envs()
for env in envs:
experiment.set_env(env)
if config.get_env_type() != 'unity':
max_steps = 128
max_episode_steps = 2
scores = experiment.train(max_steps=max_steps,
eval_frequency=16,
eval_steps=4,
max_episode_steps=max_episode_steps)
assert len(scores) == max_steps / max_episode_steps