def breakout_env_test():
env = wrap_env(gym.make('Breakout-v0'))
obs = env.reset()
counter = 0
while True:
counter += 1
env.render()
obs, _, done, _ = env.step(np.random.randint(env.action_space.n))
if done:
print('counter = ', counter)
break
def show_result():
env = wrap_env(gym.make("BreakoutNoFrameskip-v4"))
act = simple.ActWrapper.load("breakout_model.ckpt", num_cpus=1)
while True:
obs, done = env.reset(), False
episode_rew = 0
while not done:
env.render()
obs, rew, done, _ = env.step(act(obs[None])[0])
episode_rew += rew
print("Episode reward", episode_rew)
def main():
env = wrap_env(gym.make("PongNoFrameskip-v4"))
act = simple.learn(env,
q_func_pong,
n_steps=2000000,
exploration_fraction=0.20,
final_epsilon=0.01,
alpha=1e-3,
buffer_size=10000,
train_main_every=4,
update_target_every=1000,
gamma=0.99,
print_every=1,
pre_run_steps=10000,
callback=callback)
# show_result(env, act)
act.save("./pong_model.ckpt")
def main():
env = wrap_env(gym.make("BreakoutNoFrameskip-v4"))
n_steps = 500000
act = simple.learn(env,
q_func_breakout,
n_steps=n_steps,
exploration_fraction=0.2,
final_epsilon=0.01,
alpha=5e-4,
buffer_size=10000,
train_main_every=4,
update_target_every=1000,
gamma=0.99,
print_every=4,
pre_run_steps=10000,
callback=callback)
# show_result(env, act)
act.save("./breakout_model.ckpt")