memory=memory,
nb_steps_warmup=100,
enable_dueling_network=True,
dueling_type='avg',
target_model_update=1e-3,
policy=policy)
dqn.compile(Adam(lr=1e-4), metrics=['mae'])
# Okay, now it's time to learn something! We visualize the training here for show, but this
# slows down training quite a lot. You can always safely abort the training prematurely using
# Ctrl + C.
history = dqn.learning(env,
Given_policy,
policy_list,
nb_steps=5e6,
visualize=False,
log_interval=1000,
verbose=2,
nb_max_episode_steps=1000,
imitation_leaning_time=0,
reinforcement_learning_time=1e10)
sio.savemat(ENV_NAME + '-' + nowtime + '/fit.mat', history.history)
# After training is done, we save the final weights.
dqn.save_weights(ENV_NAME + '-' + nowtime + '/fit-weights.h5f', overwrite=True)
# Finally, evaluate our algorithm for 5 episodes.
history = dqn.test(env,
nb_episodes=10,
visualize=True,
nb_max_episode_steps=5000)
sio.savemat(ENV_NAME + '-' + nowtime + '/test.mat', history.history)
