def watch(sess_):
with sess_.as_default():
env_monitor = GymEnvironment(env_name=env_name,
log_path=None,
render=True)
info_ = env_monitor.reset()[brain_name]
trainer_monitor = Trainer(ppo_model, sess_, info_, is_continuous,
use_observations, use_states, False)
steps_ = sess_.run([ppo_model.global_step])
print("Starting watcher.")
while True:
done = False
info_ = env_monitor.reset()[brain_name]
while not done:
info_ = trainer_monitor.take_action(info_,
env_monitor,
brain_name,
steps_,
normalize,
stochastic=False)
done = info_.local_done[0]
init = tf.global_variables_initializer()
saver = tf.train.Saver(max_to_keep=keep_checkpoints)
with tf.Session() as sess:
# Instantiate model parameters
if load_model:
print('Loading Model...')
ckpt = tf.train.get_checkpoint_state(model_path)
saver.restore(sess, ckpt.model_checkpoint_path)
else:
sess.run(init)
steps = sess.run(ppo_model.global_step)
summary_writer = tf.summary.FileWriter(summary_path)
info = env.reset(train_mode=train_model)[brain_name]
trainer = Trainer(ppo_model, sess, info, is_continuous, use_observations,
use_states)
while steps <= max_steps or not train_model:
if env.global_done:
info = env.reset(train_mode=train_model)[brain_name]
# Decide and take an action
new_info = trainer.take_action(info, env, brain_name)
info = new_info
trainer.process_experiences(info, time_horizon, gamma, lambd)
if len(trainer.training_buffer['actions']
) > buffer_size and train_model:
# Perform gradient descent with experience buffer
trainer.update_model(batch_size, num_epoch)
if steps % summary_freq == 0 and steps != 0 and train_model:
# Write training statistics to tensorboard.
trainer.write_summary(summary_writer, steps)
if steps % save_freq == 0 and steps != 0 and train_model:
with tf.Session() as sess:
# Instantiate model parameters
if load_model:
print('Loading Model...')
ckpt = tf.train.get_checkpoint_state(model_path)
if ckpt == None:
print('The model {0} could not be found. Make sure you specified the right '
'--run-path'.format(model_path))
saver.restore(sess, ckpt.model_checkpoint_path)
else:
sess.run(init)
steps, last_reward = sess.run([ppo_model.global_step, ppo_model.last_reward])
summary_writer = tf.summary.FileWriter(summary_path)
info = env.reset(train_mode=train_model, progress=get_progress())[brain_name]
trainer = Trainer(ppo_model, sess, info, is_continuous, use_observations, use_states, train_model)
if train_model:
trainer.write_text(summary_writer, 'Hyperparameters', options, steps)
while steps <= max_steps or not train_model:
if env.global_done:
info = env.reset(train_mode=train_model, progress=get_progress())[brain_name]
trainer.reset_buffers(info, total=True)
# Decide and take an action
new_info = trainer.take_action(info, env, brain_name, steps, normalize)
info = new_info
trainer.process_experiences(info, time_horizon, gamma, lambd)
if len(trainer.training_buffer['actions']) > buffer_size and train_model:
# Perform gradient descent with experience buffer
trainer.update_model(batch_size, num_epoch)
if steps % summary_freq == 0 and steps != 0 and train_model:
# Write training statistics to tensorboard.
if load_model:
print('Loading Model...')
ckpt = tf.train.get_checkpoint_state(model_path)
if ckpt is None:
print(
'The model {0} could not be found. Make sure you specified the right --run-path'
.format(model_path))
saver.restore(sess, ckpt.model_checkpoint_path)
else:
sess.run(init)
steps, last_reward = sess.run(
[ppo_model.global_step, ppo_model.last_reward])
summary_writer = tf.summary.FileWriter(summary_path)
info = env.reset()[brain_name]
trainer = Trainer(ppo_model, sess, info, is_continuous, use_observations,
use_states, train_model)
trainer_monitor = Trainer(ppo_model, sess, info, is_continuous,
use_observations, use_states, False)
render_started = False
while steps <= max_steps or not train_model:
if env.global_done:
info = env.reset()[brain_name]
trainer.reset_buffers(info, total=True)
# Decide and take an action
info = trainer.take_action(info,
env,
brain_name,
steps,
normalize_steps,
stochastic=True)