class EnvWrapper(object):
'''
Wrapper around UnityEnvironment that resets each arena if the episode is done
It will only work correctly if using a single arena on each environment
'''
def __init__(self, *args, **kwargs):
'''
Check UnityEnvironment parameters
'''
self._env = UnityEnvironment(*args, **kwargs)
self._arenas_configurations = None
def __getattr__(self, attr):
if attr in self.__dict__:
return getattr(self, attr)
return getattr(self._env, attr)
def reset(self, arenas_configurations=None, train_mode=True):
""" Shuffle arenas and reset configuration """
if arenas_configurations is not None:
self._arenas_configurations = arenas_configurations
self._arenas_configurations.shuffle_arenas()
return self._env.reset(self._arenas_configurations, train_mode)
def step(self, *args, **kwargs):
ret = self._env.step(*args, **kwargs)
if ret['Learner'].local_done[0]:
new_ret = self.reset()
ret['Learner'].visual_observations = new_ret['Learner'].visual_observations
return ret
def init_environment(env_path, docker_target_name, no_graphics, worker_id,
seed):
if env_path is not None:
# Strip out executable extensions if passed
env_path = (env_path.strip().replace('.app', '').replace(
'.exe', '').replace('.x86_64', '').replace('.x86', ''))
docker_training = docker_target_name is not None
return UnityEnvironment(n_arenas=1,
file_name=env_path,
worker_id=worker_id,
seed=seed,
docker_training=docker_training,
play=True)
def debug_confirm_arena_config(env_path, arena_config):
env_path = (env_path.strip().replace('.app', '').replace(
'.exe', '').replace('.x86_64', '').replace('.x86', ''))
from animalai.envs.environment import UnityEnvironment
env = UnityEnvironment(n_arenas=16,
file_name=env_path,
worker_id=1,
seed=0,
docker_training=False,
play=True)
env.reset(arenas_configurations=arena_config)
try:
while True:
continue
except KeyboardInterrupt:
env.close()
def main(args):
docker_training = docker_target_name is not None
env = UnityEnvironment(
n_arenas=args.n_arenas,
file_name=env_path,
worker_id=worker_id,
seed=seed,
docker_training=docker_training,
play=False,
resolution=resolution
)
arena_config_in = ArenaConfig('configs/3-Obstacles.yaml')
env.reset(arenas_configurations=arena_config_in)
start_time = time.time()
for i in range(args.frames):
res = env.step(np.random.randint(0, 3, size=2 * args.n_arenas))
elapsed_time = time.time() - start_time
fps = float(args.frames) / elapsed_time
print("n_arenas={0}, fps={1:.3f}".format(args.n_arenas, fps))
env.close()
resolution = 84
n_channels = 3
dim_actions = 2
sample_size_per_task = 30
# set up the testing env
np.random.seed(seed)
if env_path is not None:
env_path = (env_path.strip().replace('.app', '').replace(
'.exe', '').replace('.x86_64', '').replace('.x86', ''))
docker_training = docker_target_name is not None
env = UnityEnvironment(n_arenas=n_arenas,
file_name=env_path,
worker_id=worker_id,
seed=seed,
docker_training=docker_training,
play=False,
resolution=resolution)
# The Agent to test
agent = Agent()
# visualization
plt.ion()
fig, ax = plt.subplots(ncols=1, nrows=1)
image = ax.imshow(np.zeros(
(resolution, resolution, 3))) # visual input for the agent
line, = ax.plot(
[], []) # the direction that the agent want to go given the visual input
sca = ax.scatter([], [], s=5, c="yellow") # the path to the target(food)
def __init__(self, *args, **kwargs):
'''
Check UnityEnvironment parameters
'''
self._env = UnityEnvironment(*args, **kwargs)
self._arenas_configurations = None
from animalai.envs.environment import UnityEnvironment
from animalai.envs.arena_config import ArenaConfig
import time
import random
worker_id = random.randint(0, 100)
env = UnityEnvironment(n_arenas=1,
file_name='../env_modified/AnimalAILinux',
worker_id=worker_id,
seed=0,
docker_training=False,
inference=True)
arena_config_in = ArenaConfig('configs/empty.yaml')
env.reset(arenas_configurations=arena_config_in)
total_distance = 0
# step_time_length = 0.0595
try:
while True:
step = 0
direction = input()
if direction == "w":
action = [1, 0]
if direction == "s":
action = [2, 0]
if direction == "d":
action = [0, 1]
if direction == "a":
action = [0, 2]
while step < 1.0: