def discrete_atari_env():
env = AtariEnvironment(name="MsPacman-v0",
clone_seeds=True,
autoreset=True)
env.reset()
env = DiscreteEnv(env)
return env
def __init__(self, swarm, *args, **kwargs):
super(DistributedRam, self).__init__(swarm=swarm, *args, **kwargs)
self.local_swarm = swarm()
env = self.local_swarm.env
env_name = env.name if isinstance(
env, ParallelEnvironment) else env._env.name
self.local_env = AtariEnvironment(name=env_name, clone_seeds=True)
self.local_env.reset()
def atari_environment():
game_name = "MsPacman-v0"
plangym_env = AtariEnvironment(
name=game_name,
clone_seeds=True,
autoreset=True,
)
return DiscreteEnv(env=plangym_env)
def atari_env():
env = AtariEnvironment(name="MsPacman-v0",
clone_seeds=True,
autoreset=True)
env.reset()
env = DiscreteEnv(env)
params = {
"actions": {
"dtype": np.int64
},
"critic": {
"dtype": np.float32
}
}
states = States(state_dict=params, batch_size=N_WALKERS)
states.update(actions=np.ones(N_WALKERS), critic=np.ones(N_WALKERS))
return env, states
def create_atari_swarm():
env = AtariEnvironment(name="MsPacman-ram-v0", )
dt = GaussianDt(min_dt=10, max_dt=100, loc_dt=5, scale_dt=2)
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x, critic=dt),
env=lambda: DiscreteEnv(env),
n_walkers=6,
max_epochs=10,
reward_scale=2,
reward_limit=1,
)
return swarm
def create_atari_swarm():
env = AtariEnvironment(name="MsPacman-ram-v0",
clone_seeds=True,
autoreset=True)
dt = GaussianDt(min_dt=3, max_dt=100, loc_dt=5, scale_dt=2)
swarm = Swarm(
model=lambda x: DiscreteUniform(env=x, critic=dt),
walkers=Walkers,
env=lambda: DiscreteEnv(env),
n_walkers=67,
max_epochs=500,
reward_scale=2,
reward_limit=751,
)
return swarm
def create_step_to_best_after_impr():
from plangym import AtariEnvironment
from fragile.core import GaussianDt
env = AtariEnvironment(name="MsPacman-ram-v0", clone_seeds=True, autoreset=True)
dt = GaussianDt(min_dt=3, max_dt=100, loc_dt=5, scale_dt=2)
swarm = StepToBest(
model=lambda x: DiscreteUniform(env=x, critic=dt),
env=lambda: DiscreteEnv(env),
reward_limit=-100,
n_walkers=67,
max_epochs=60,
step_epochs=5,
step_after_improvement=True,
)
return swarm
class DistributedRam(DistributedSwarm):
def __init__(self, swarm, *args, **kwargs):
super(DistributedRam, self).__init__(swarm=swarm, *args, **kwargs)
self.local_swarm = swarm()
env = self.local_swarm.env
env_name = env.name if isinstance(
env, ParallelEnvironment) else env._env.name
self.local_env = AtariEnvironment(name=env_name, clone_seeds=True)
self.local_env.reset()
def image_from_state(self, state):
self.local_env.set_state(state.astype(np.uint8).copy())
self.local_env.step(0)
return np.asarray(self.local_env._env.ale.getScreenRGB(),
dtype=np.uint8)
def stream_progress(self, state, observation, reward):
example = pd.DataFrame({"reward": [reward]},
index=[self.n_iters // self.n_swarms])
self.stream.emit(example)
obs = self.image_from_state(state)
self.frame_pipe.send(obs)
def qbert_rgb():
env = AtariEnvironment(name="Qbert-v0", clone_seeds=True, autoreset=True)
env.reset()
env = AtariEnv(env)
return env
def pacman_ram():
env = AtariEnvironment(name="MsPacman-ram-v0", clone_seeds=True, autoreset=True)
env.reset()
env = AtariEnv(env)
return env