def make_env(i):
env = gym.make(args.env)
env = MultiMonitor(env,
os.path.join(out_path, 'mon{:d}'.format(i)),
allow_early_resets=True)
env.seed(args.seed + 10 * i)
return env
def test_multi_monitor():
"""Smoke test for MultiMonitor."""
env = gym.make("aprl/IteratedMatchingPennies-v0")
env.seed(42)
with tempfile.TemporaryDirectory(prefix="test_multi_mon") as d:
env = MultiMonitor(env, filename=os.path.join(d, "test"))
for eps in range(5):
env.reset()
done = False
while not done:
a = env.action_space.sample()
_, _, done, info = env.step(a)
epinfo = info["episode"]
assert set(epinfo.keys()) == {"r", "r0", "r1", "l", "t"}
def test_multi_monitor():
"""Smoke test for MultiMonitor."""
env = gym.make('aprl/IteratedMatchingPennies-v0')
env.seed(42)
with tempfile.TemporaryDirectory(prefix='test_multi_mon') as d:
env = MultiMonitor(env, filename=os.path.join(d, 'test'))
for eps in range(5):
env.reset()
done = False
while not done:
a = env.action_space.sample()
_, _, done, info = env.step(a)
epinfo = info['episode']
assert set(epinfo.keys()) == {'r', 'r0', 'r1', 'l', 't'}
def make_env(env_name,
seed,
i,
out_dir,
our_idx=None,
pre_wrapper=None,
post_wrapper=None,
agent_wrappers=None):
multi_env = gym.make(env_name)
if agent_wrappers is not None:
for agent_id in agent_wrappers:
multi_env.agents[agent_id] = agent_wrappers[agent_id](
multi_env.agents[agent_id])
if pre_wrapper is not None:
multi_env = pre_wrapper(multi_env)
if not isinstance(multi_env, MultiAgentEnv):
multi_env = SingleToMulti(multi_env)
multi_env.seed(seed + i)
if out_dir is not None:
mon_dir = osp.join(out_dir, 'mon')
os.makedirs(mon_dir, exist_ok=True)
multi_env = MultiMonitor(multi_env,
osp.join(mon_dir, 'log{}'.format(i)), our_idx)
if post_wrapper is not None:
multi_env = post_wrapper(multi_env)
return multi_env
def make_env(
env_name,
seed,
i,
out_dir,
our_idx=None,
pre_wrappers=None,
post_wrappers=None,
agent_wrappers=None,
):
multi_env = gym.make(env_name)
if agent_wrappers is not None:
for agent_id in agent_wrappers:
multi_env.agents[agent_id] = agent_wrappers[agent_id](
multi_env.agents[agent_id])
multi_env = _apply_wrappers(pre_wrappers, multi_env)
if not hasattr(multi_env, "num_agents"):
multi_env = SingleToMulti(multi_env)
if out_dir is not None:
mon_dir = osp.join(out_dir, "mon")
os.makedirs(mon_dir, exist_ok=True)
multi_env = MultiMonitor(multi_env,
osp.join(mon_dir, "log{}".format(i)), our_idx)
multi_env = _apply_wrappers(post_wrappers, multi_env)
multi_env.seed(seed + i)
return multi_env