def build_visualizer(self, record_path):
atari_params = {'environment': self._environment}
atari_plot = atari_plotter.AtariPlotter(parameter_dict=atari_params)
return agent_visualizer.AgentVisualizer(
record_path=record_path,
plotters=[atari_plot],
screen_width=atari_plot.parameters['width'],
screen_height=atari_plot.parameters['height'])
def visualize(self, record_path, num_global_steps=500):
if not tf.io.gfile.exists(record_path):
tf.io.gfile.makedirs(record_path)
self._agent.eval_mode = True
# Set up the game playback rendering.
atari_params = {'environment': self._environment}
atari_plot = atari_plotter.AtariPlotter(parameter_dict=atari_params)
# Plot the rewards received next to it.
reward_params = {
'x': atari_plot.parameters['width'],
'xlabel': 'Timestep',
'ylabel': 'Reward',
'title': 'Rewards',
'get_line_data_fn': self._agent.get_rewards
}
reward_plot = line_plotter.LinePlotter(parameter_dict=reward_params)
action_names = [
'Action {}'.format(x) for x in range(self._agent.num_actions)
]
# Plot Q-values (DQN) or Q-value distributions (Rainbow).
q_params = {
'x': atari_plot.parameters['width'] // 2,
'y': atari_plot.parameters['height'],
'legend': action_names
}
if 'DQN' in self._agent.__class__.__name__:
q_params['xlabel'] = 'Timestep'
q_params['ylabel'] = 'Q-Value'
q_params['title'] = 'Q-Values'
q_params['get_line_data_fn'] = self._agent.get_q_values
q_plot = line_plotter.LinePlotter(parameter_dict=q_params)
elif 'Implicit' in self._agent.__class__.__name__:
q_params['xlabel'] = 'Timestep'
q_params['ylabel'] = 'Quantile Value'
q_params['title'] = 'Quantile Values'
q_params['get_line_data_fn'] = self._agent.get_q_values
q_plot = line_plotter.LinePlotter(parameter_dict=q_params)
else:
q_params['xlabel'] = 'Return'
q_params['ylabel'] = 'Return probability'
q_params['title'] = 'Return distribution'
q_params['get_bar_data_fn'] = self._agent.get_probabilities
q_plot = bar_plotter.BarPlotter(parameter_dict=q_params)
screen_width = (atari_plot.parameters['width'] +
reward_plot.parameters['width'])
screen_height = (atari_plot.parameters['height'] +
q_plot.parameters['height'])
# Dimensions need to be divisible by 2:
if screen_width % 2 > 0:
screen_width += 1
if screen_height % 2 > 0:
screen_height += 1
visualizer = agent_visualizer.AgentVisualizer(
record_path=record_path,
plotters=[atari_plot, reward_plot, q_plot],
screen_width=screen_width,
screen_height=screen_height)
global_step = 0
while global_step < num_global_steps:
initial_observation = self._environment.reset()
action = self._agent.begin_episode(initial_observation)
while True:
observation, reward, is_terminal, _ = self._environment.step(
action)
global_step += 1
visualizer.visualize()
if self._environment.game_over or global_step >= num_global_steps:
break
elif is_terminal:
self._agent.end_episode(reward)
action = self._agent.begin_episode(observation)
else:
action = self._agent.step(reward, observation)
self._end_episode(reward)
visualizer.generate_video()
def visualize(self, record_path, num_global_steps=500):
'''customize viz for bubble
- origin from MyRunner.visualize()
'''
print('RUN> visualize(%s, %d)' % (record_path, num_global_steps))
if not tf.gfile.Exists(record_path):
tf.gfile.MakeDirs(record_path)
self._agent.eval_mode = True
# Set up the game playback rendering.
atari_params = {
'environment': self._environment,
'width': 240,
'height': 224
}
atari_plot = atari_plotter.AtariPlotter(parameter_dict=atari_params)
# Plot the rewards received next to it.
reward_params = {
'x': atari_plot.parameters['width'],
'xlabel': 'Timestep',
'ylabel': 'Reward',
'title': 'Rewards',
'get_line_data_fn': self._agent.get_rewards
}
#reward_plot = line_plotter.LinePlotter(parameter_dict=reward_params)
reward_plot = MyLinePlotter(parameter_dict=reward_params)
action_names = [
'Action {}'.format(x) for x in range(self._agent.num_actions)
]
# Plot Observation at left-bottom
obsrv_params = {
'x': atari_plot.parameters['x'],
'y': atari_plot.parameters['height'] - 10,
'width': atari_plot.parameters['width'],
'height': atari_plot.parameters['height'],
}
obsrv_plot = MyObservationPlotter(parameter_dict=obsrv_params)
# Plot Q-values (DQN) or Q-value distributions (Rainbow).
q_params = {
'x': atari_plot.parameters['width'],
'y': atari_plot.parameters['height'],
'legend': action_names
}
if 'DQN' in self._agent.__class__.__name__:
q_params['xlabel'] = 'Timestep'
q_params['ylabel'] = 'Q-Value'
q_params['title'] = 'Q-Values'
q_params['get_line_data_fn'] = self._agent.get_q_values
q_plot = MyLinePlotter(parameter_dict=q_params)
else:
q_params['xlabel'] = 'Return'
q_params['ylabel'] = 'Return probability'
q_params['title'] = 'Return distribution'
q_params['get_bar_data_fn'] = self._agent.get_probabilities
q_plot = MyBarPlotter(parameter_dict=q_params)
# Screen Size
screen_width = (atari_plot.parameters['width'] +
reward_plot.parameters['width'])
screen_height = (atari_plot.parameters['height'] +
q_plot.parameters['height'])
# Dimensions need to be divisible by 2:
screen_width += 1 if screen_width % 2 > 0 else 0
screen_height += 1 if screen_height % 2 > 0 else 0
# build visualizer.
visualizer = agent_visualizer.AgentVisualizer(
record_path=record_path,
plotters=[atari_plot, reward_plot, obsrv_plot, q_plot],
screen_width=screen_width,
screen_height=screen_height)
# run loop in global_step
global_step = 0
while global_step < num_global_steps:
initial_observation = self._environment.reset()
action = self._agent.begin_episode(initial_observation)
while True:
observation, reward, is_terminal, info = self._environment.step(
action)
global_step += 1
obsrv_plot.setObservation(observation)
visualizer.visualize()
if self._environment.game_over or global_step >= num_global_steps:
break
elif is_terminal:
self._agent.end_episode(reward)
action = self._agent.begin_episode(observation)
else:
action = self._agent.step(reward, observation, info)
self._end_episode(reward)
visualizer.generate_video()