Parallelization of this "CartPole v0" standard environment failed

[fengyuhun]

import tensorflow as tf
from tf_agents.environments import suite_gym
from tf_agents.environments import tf_py_environment
from tf_agents.environments import parallel_py_environment

创建单个CartPole环境

def get_CartPole_Env():
return suite_gym.load("CartPole-v0")

创建并行环境

tf_env = tf_py_environment.TFPyEnvironment(
parallel_py_environment.ParallelPyEnvironment([get_CartPole_Env] * 4)
)

RuntimeError Traceback (most recent call last)
in <cell line: 11>()
10 # 创建并行环境
11 tf_env = tf_py_environment.TFPyEnvironment(
---> 12 parallel_py_environment.ParallelPyEnvironment([get_CartPole_Env] * 4)
13 )

6 frames
/usr/local/lib/python3.10/dist-packages/tf_agents/system/system_multiprocessing.py in get_context(method)
85 """
86 if not multiprocessing_core.initialized():
---> 87 raise RuntimeError(_NOT_INITIALIZED_ERROR)
88 return _rewrite_target_with_state(multiprocessing_core.get_context(method))
89

RuntimeError: Unable to load multiprocessing context.

Please ensure that you properly initialize your program by wrapping your main()
call:

def main(argv):
...

if name == 'main':
tf_agents.system.multiprocessing.handle_main(main, extra_state_savers=...)

or, if using absl.app:

if name == 'main':
tf_agents.system.multiprocessing.handle_main(
functools.partial(absl.app.run, main), extra_state_savers=...)

For unit tests, this also means wrapping your test.main using handle_test_main:

if name == 'main':
tf_agents.system.multiprocessing.handle_test_main(
tf.test.main, extra_state_savers=...)

or

if name == 'main':
tf_agents.system.multiprocessing.handle_test_main(
tf_agents.utils.test_utils.main, extra_state_savers=...)

If you are in interactive mode (e.g. python console, ipython, jupyter notebook)
use:

tf_agents.system.multiprocessing.enable_interactive_mode(
extra_state_savers=...)

For more details on state savers, see the docstrings for
tf_agents.multiprocessing.handle_* and:

https://pythonspeed.com/articles/python-multiprocessing/

In call to configurable 'ParallelPyEnvironment' (<class 'tf_agents.environments.parallel_py_environment.ParallelPyEnvironment'>)