1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102 | import time
from typing import Optional
import warnings
import numpy as np
from lit_sb3.vec_env.base_vec_env import VecEnv, VecEnvObs, VecEnvStepReturn, VecEnvWrapper
class VecMonitor(VecEnvWrapper):
"""A vectorized monitor wrapper for *vectorized* Gym environments,.
it is used to record the episode reward, length, time and other data.
Some environments like `openai/procgen <https://github.com/openai/procgen>`_
or `gym3 <https://github.com/openai/gym3>`_ directly initialize the
vectorized environments, without giving us a chance to use the ``Monitor``
wrapper. So this class simply does the job of the ``Monitor`` wrapper on
a vectorized level.
Args:
venv: The vectorized environment
filename: the location to save a log file, can be None for no log
info_keywords: extra information to log, from the information return of env.step()
"""
def __init__(
self,
venv: VecEnv,
filename: str | None = None,
info_keywords: tuple[str, ...] = (),
) -> None:
# Avoid circular import
from lit_sb3.wrappers.monitor import Monitor, ResultsWriter
# This check is not valid for special `VecEnv`
# like the ones created by Procgen, that does follow completely
# the `VecEnv` interface
try:
is_wrapped_with_monitor = venv.env_is_wrapped(Monitor)[0]
except AttributeError:
is_wrapped_with_monitor = False
if is_wrapped_with_monitor:
warnings.warn(
"The environment is already wrapped with a `Monitor` wrapper"
"but you are wrapping it with a `VecMonitor` wrapper, the `Monitor` statistics will be"
"overwritten by the `VecMonitor` ones.",
UserWarning,
stacklevel=2,
)
VecEnvWrapper.__init__(self, venv)
self.episode_count = 0
self.t_start = time.time()
env_id = None
if hasattr(venv, "spec") and venv.spec is not None:
env_id = venv.spec.id # ty:ignore[unresolved-attribute]
self.results_writer: ResultsWriter | None = None
if filename:
self.results_writer = ResultsWriter(
filename, header={"t_start": self.t_start, "env_id": str(env_id)}, extra_keys=info_keywords
)
self.info_keywords = info_keywords
self.episode_returns = np.zeros(self.num_envs, dtype=np.float32)
self.episode_lengths = np.zeros(self.num_envs, dtype=np.int32)
def reset(self) -> VecEnvObs:
obs = self.venv.reset()
self.episode_returns = np.zeros(self.num_envs, dtype=np.float32)
self.episode_lengths = np.zeros(self.num_envs, dtype=np.int32)
return obs
def step_wait(self) -> VecEnvStepReturn:
obs, rewards, dones, infos = self.venv.step_wait()
self.episode_returns += rewards
self.episode_lengths += 1
new_infos = list(infos[:])
for i in range(len(dones)):
if dones[i]:
info = infos[i].copy()
episode_return = self.episode_returns[i]
episode_length = self.episode_lengths[i]
episode_info = {"r": episode_return, "l": episode_length, "t": round(time.time() - self.t_start, 6)}
for key in self.info_keywords:
episode_info[key] = info[key]
info["episode"] = episode_info
self.episode_count += 1
self.episode_returns[i] = 0
self.episode_lengths[i] = 0
if self.results_writer:
self.results_writer.write_row(episode_info)
new_infos[i] = info
return obs, rewards, dones, new_infos
def close(self) -> None:
if self.results_writer:
self.results_writer.close()
return self.venv.close()
|