def test_env_runner_log_data_interaction_debug_log(mock_data_logger, mock_task,
mock_agent):
# Assign
mock_task.step.return_value = ([1, 0.1], -1, False, {})
mock_agent.act.return_value = 1
env_runner = EnvRunner(mock_task,
mock_agent,
data_logger=mock_data_logger,
debug_log=True)
# Act
env_runner.interact_episode(eps=0.1,
max_iterations=10,
log_interaction_freq=None)
env_runner.log_data_interaction()
# Assert
mock_agent.log_metrics.assert_called_once_with(mock_data_logger,
10,
full_log=False)
assert mock_data_logger.log_values_dict.call_count == 20 # 10x iter per states and actions
assert mock_data_logger.log_value.call_count == 20 # 10x iter per rewards and dones
task: TaskType = GymTask(env_name)
config = {
'rollout_length': 30,
'batch_size': 30,
"number_updates": 1,
"ppo_ratio_clip": 0.2,
"value_loss_weight": 2,
"entropy_weight": 0.0005,
"gamma": 0.98,
"action_scale": 2,
"max_grad_norm_actor": 2.0,
"max_grad_norm_critic": 2.0,
"critic_lr": 1e-3,
"actor_lr": 1e-3,
}
agent = Agent(task.state_size, task.action_size, hidden_layers=(300, 300), config=config)
env_runner = EnvRunner(task, agent)
env_runner.interact_episode(0, render=True)
scores = env_runner.run(80, 4000)
env_runner.interact_episode(0, render=True)
# plot the scores
fig = plt.figure()
ax = fig.add_subplot(111)
plt.plot(np.arange(len(scores)), scores)
plt.ylabel('Score')
plt.xlabel('Episode #')
plt.savefig(f'{env_name}.png', dpi=120)
plt.show()