def create_doom(env_id,
client_id,
envWrap=True,
record=False,
outdir=None,
noLifeReward=False,
acRepeat=0,
**_):
from ppaquette_gym_doom import wrappers
if 'labyrinth' in env_id.lower():
if 'single' in env_id.lower():
env_id = 'ppaquette/LabyrinthSingle-v0'
elif 'fix' in env_id.lower():
env_id = 'ppaquette/LabyrinthManyFixed-v0'
else:
env_id = 'ppaquette/LabyrinthMany-v0'
elif 'very' in env_id.lower():
env_id = 'ppaquette/DoomMyWayHomeFixed15-v0'
elif 'sparse' in env_id.lower():
env_id = 'ppaquette/DoomMyWayHomeFixed-v0'
elif 'fix' in env_id.lower():
if '1' in env_id or '2' in env_id:
env_id = 'ppaquette/DoomMyWayHomeFixed' + str(env_id[-2:]) + '-v0'
elif 'new' in env_id.lower():
env_id = 'ppaquette/DoomMyWayHomeFixedNew-v0'
else:
env_id = 'ppaquette/DoomMyWayHomeFixed-v0'
else:
env_id = 'ppaquette/DoomMyWayHome-v0'
# VizDoom workaround: Simultaneously launching multiple vizdoom processes
# makes program stuck, so use the global lock in multi-threading/processing
client_id = int(client_id)
time.sleep(client_id * 10)
env = gym.make(env_id)
modewrapper = wrappers.SetPlayingMode('algo')
obwrapper = wrappers.SetResolution('160x120')
acwrapper = wrappers.ToDiscrete('minimal')
env = modewrapper(obwrapper(acwrapper(env)))
# env = env_wrapper.MakeEnvDynamic(env) # to add stochasticity
if record and outdir is not None:
env = gym.wrappers.Monitor(env, outdir, force=True)
if envWrap:
fshape = (42, 42)
frame_skip = acRepeat if acRepeat > 0 else 4
env.seed(None)
if noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = env_wrapper.BufferedObsEnv(env, skip=frame_skip, shape=fshape)
env = env_wrapper.SkipEnv(env, skip=frame_skip)
elif noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = Vectorize(env)
env = DiagnosticsInfo(env)
env = Unvectorize(env)
return env
def __init__(self, level='ppaquette/DoomMyWayHome-v0'):
time.sleep(np.random.randint(100))
env = gym.make(level)
modewrapper = wrappers.SetPlayingMode('algo')
obwrapper = wrappers.SetResolution('160x120')
acwrapper = wrappers.ToDiscrete('minimal')
env = modewrapper(obwrapper(acwrapper(env)))
frame_skip = options.get('environment/frame_skip', None)
if frame_skip is not None:
skip_wrapper = SkipWrapper(frame_skip)
env = skip_wrapper(env)
self._record = options.get('environment/record', False)
if self._record:
out_dir = options.get('environment/out_dir',
'/tmp/' + level.split('/')[-1])
if not os.path.exists(out_dir):
os.makedirs(out_dir)
env = gym.wrappers.Monitor(env, out_dir, force=True)
self._no_op_max = options.get('environment/no_op_max', 0)
self._reset_action = env.action_space.sample() \
if options.get('environment/stochastic_reset', False) else 0
env.seed(random.randrange(1000000))
self._show_ui = options.get('show_ui', False)
limit = options.get(
'environment/limit',
env.spec.tags.get('wrapper_config.TimeLimit.max_episode_steps'))
if limit is not None:
env._max_episode_steps = limit
shape = options.get('environment/shape', (42, 42))
self._shape = shape[:2]
self._channels = 0 if len(shape) == 2 else shape[-1]
self.action_size = self._get_action_size(env)
if self.action_size != options.algorithm.output.action_size:
print(
'Algorithm expects different action size (%d) from gym (%d). \n'
'Please set correct action size in you configuration yaml.' %
(options.algorithm.output.action_size, self.action_size))
sys.exit(-1)
self.env = NoNegativeRewardEnv(env)
self._obs_buffer = deque(maxlen=2)
self.observation_space = Box(0.0, 255.0, shape)
self.observation_space.high[...] = 1.0
self._scale = (1.0 / 255.0)
self.reset()
def create_mario(env_id,
client_id,
envWrap=True,
record=False,
outdir=None,
noLifeReward=False,
acRepeat=0,
**_):
import ppaquette_gym_super_mario
from ppaquette_gym_super_mario import wrappers
if '-v' in env_id.lower():
env_id = 'ppaquette/' + env_id
else:
env_id = 'ppaquette/SuperMarioBros-1-3-v0' # shape: (224,256,3)=(h,w,c)
#env_id = 'ppaquette/SuperMarioBros-4-2-Tiles-v0'
# Mario workaround: Simultaneously launching multiple vizdoom processes makes program stuck,
# so use the global lock in multi-threading/multi-processing
# see: https://github.com/ppaquette/gym-super-mario/tree/master/ppaquette_gym_super_mario
client_id = int(client_id)
time.sleep(client_id * 50)
env = gym.make(env_id)
modewrapper = wrappers.SetPlayingMode('algo')
acwrapper = wrappers.ToDiscrete()
env = modewrapper(acwrapper(env))
env = env_wrapper.MarioEnv(env, tilesEnv=False)
#env = env_wrapper.MarioEnv(env, tilesEnv=True)
if record and outdir is not None:
env = gym.wrappers.Monitor(env, outdir, force=True)
if envWrap:
frame_skip = acRepeat if acRepeat > 0 else 4
frame_skip = 6 if "1-1" in env_id else frame_skip
fshape = (42, 42)
env.seed(None)
if noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = env_wrapper.BufferedObsEnv(env,
skip=frame_skip,
shape=fshape,
maxFrames=False)
if frame_skip > 1:
env = env_wrapper.SkipEnv(env, skip=frame_skip)
elif noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = Vectorize(env)
env = DiagnosticsInfo(env)
env = Unvectorize(env)
# env.close() # TODO: think about where to put env.close !
return env
def create_doom(record=False, outdir=None):
from ppaquette_gym_doom import wrappers
env = gym.make('ppaquette/DoomMyWayHome-v0')
modewrapper = wrappers.SetPlayingMode('algo')
obwrapper = wrappers.SetResolution('160x120')
acwrapper = wrappers.ToDiscrete('minimal')
env = modewrapper(obwrapper(acwrapper(env)))
if record:
env = gym.wrappers.Monitor(env, outdir, force=True)
fshape = (42, 42)
env.seed(None)
#env = env_wrapper.NoNegativeRewardEnv(env)
env = BufferedObsEnv(env, skip=1, shape=fshape)
return env
def create_mario(env_id,
client_id,
envWrap=True,
record=False,
outdir=None,
noLifeReward=False,
acRepeat=0,
**_):
import ppaquette_gym_super_mario
from ppaquette_gym_super_mario import wrappers
if '-v' in env_id.lower():
env_id = 'ppaquette/' + env_id
else:
env_id = 'ppaquette/SuperMarioBros-1-1-v0' # shape: (224,256,3)=(h,w,c)
# Mario workaround: Simultaneously launching multiple vizdoom processes makes program stuck,
# so use the global lock in multi-threading/multi-processing
# see: https://github.com/ppaquette/gym-super-mario/tree/master/ppaquette_gym_super_mario
client_id = int(client_id)
time.sleep(client_id * 50)
env = gym.make(env_id)
modewrapper = wrappers.SetPlayingMode('algo')
acwrapper = wrappers.ToDiscrete()
env = modewrapper(acwrapper(env))
env = env_wrapper.MarioEnv(env)
if record and outdir is not None:
env = gym.wrappers.Monitor(env, outdir, force=True)
if envWrap:
frame_skip = acRepeat if acRepeat > 0 else 6
fshape = (42, 42)
env.seed(None)
if noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = env_wrapper.BufferedObsEnv(env,
skip=frame_skip,
shape=fshape,
maxFrames=False)
if frame_skip > 1:
env = env_wrapper.SkipEnv(env, skip=frame_skip)
elif noLifeReward:
env = env_wrapper.NoNegativeRewardEnv(env)
env = Vectorize(env)
env = DiagnosticsInfo(env)
env = Unvectorize(env)
# env.close() # TODO: think about where to put env.close !
return env
# def DiagnosticsInfo(env, *args, **kwargs):
# return vectorized.VectorizeFilter(env, DiagnosticsInfoI, *args, **kwargs)
# class DiagnosticsInfoI(vectorized.Filter):
# def __init__(self, log_interval=503):
# super(DiagnosticsInfoI, self).__init__()
# self._episode_time = time.time()
# self._last_time = time.time()
# self._local_t = 0
# self._log_interval = log_interval
# self._episode_reward = 0
# self._episode_length = 0
# self._all_rewards = []
# self._num_vnc_updates = 0
# self._last_episode_id = -1
# def _after_reset(self, observation):
# logger.info('Resetting environment logs')
# self._episode_reward = 0
# self._episode_length = 0
# self._all_rewards = []
# return observation
# def _after_step(self, observation, reward, done, info):
# to_log = {}
# if self._episode_length == 0:
# self._episode_time = time.time()
# self._local_t += 1
# if info.get("stats.vnc.updates.n") is not None:
# self._num_vnc_updates += info.get("stats.vnc.updates.n")
# if self._local_t % self._log_interval == 0:
# cur_time = time.time()
# elapsed = cur_time - self._last_time
# fps = self._log_interval / elapsed
# self._last_time = cur_time
# cur_episode_id = info.get('vectorized.episode_id', 0)
# to_log["diagnostics/fps"] = fps
# if self._last_episode_id == cur_episode_id:
# to_log["diagnostics/fps_within_episode"] = fps
# self._last_episode_id = cur_episode_id
# if info.get("stats.gauges.diagnostics.lag.action") is not None:
# to_log["diagnostics/action_lag_lb"] = info["stats.gauges.diagnostics.lag.action"][0]
# to_log["diagnostics/action_lag_ub"] = info["stats.gauges.diagnostics.lag.action"][1]
# if info.get("reward.count") is not None:
# to_log["diagnostics/reward_count"] = info["reward.count"]
# if info.get("stats.gauges.diagnostics.clock_skew") is not None:
# to_log["diagnostics/clock_skew_lb"] = info["stats.gauges.diagnostics.clock_skew"][0]
# to_log["diagnostics/clock_skew_ub"] = info["stats.gauges.diagnostics.clock_skew"][1]
# if info.get("stats.gauges.diagnostics.lag.observation") is not None:
# to_log["diagnostics/observation_lag_lb"] = info["stats.gauges.diagnostics.lag.observation"][0]
# to_log["diagnostics/observation_lag_ub"] = info["stats.gauges.diagnostics.lag.observation"][1]
# if info.get("stats.vnc.updates.n") is not None:
# to_log["diagnostics/vnc_updates_n"] = info["stats.vnc.updates.n"]
# to_log["diagnostics/vnc_updates_n_ps"] = self._num_vnc_updates / elapsed
# self._num_vnc_updates = 0
# if info.get("stats.vnc.updates.bytes") is not None:
# to_log["diagnostics/vnc_updates_bytes"] = info["stats.vnc.updates.bytes"]
# if info.get("stats.vnc.updates.pixels") is not None:
# to_log["diagnostics/vnc_updates_pixels"] = info["stats.vnc.updates.pixels"]
# if info.get("stats.vnc.updates.rectangles") is not None:
# to_log["diagnostics/vnc_updates_rectangles"] = info["stats.vnc.updates.rectangles"]
# if info.get("env_status.state_id") is not None:
# to_log["diagnostics/env_state_id"] = info["env_status.state_id"]
# if reward is not None:
# self._episode_reward += reward
# if observation is not None:
# self._episode_length += 1
# self._all_rewards.append(reward)
# if done:
# logger.info('True Game terminating: env_episode_reward=%s episode_length=%s', self._episode_reward, self._episode_length)
# total_time = time.time() - self._episode_time
# to_log["global/episode_reward"] = self._episode_reward
# to_log["global/episode_length"] = self._episode_length
# to_log["global/episode_time"] = total_time
# to_log["global/reward_per_time"] = self._episode_reward / total_time
# self._episode_reward = 0
# self._episode_length = 0
# self._all_rewards = []
# if 'distance' in info: to_log['distance'] = info['distance'] # mario
# if 'POSITION_X' in info: # doom
# to_log['POSITION_X'] = info['POSITION_X']
# to_log['POSITION_Y'] = info['POSITION_Y']
# return observation, reward, done, to_log