[BUG] Action Space seeding doesn't work in some cases

[nico-bohlinger]

Describe the bug

Setting the seed in the action space only seems to work when copying the action space to a variable. Which seems like really odd behavior. I would also expect that setting the seed in envpool.make would be enough.

To Reproduce

import envpool


actions = []
for i in range(2):
    env = envpool.make("Humanoid-v4", env_type="gym", num_envs=1, seed=1)
    action_space = env.action_space
    action_space.seed(1)
    action = action_space.sample()
    actions.append(action)
print(actions[0] == actions[1])
# Outputs:
# [ True  True  True  True  True  True  True  True  True  True  True  True
# True  True  True  True  True]

actions = []
for i in range(2):
    env = envpool.make("Humanoid-v4", env_type="gym", num_envs=1, seed=1)
    env.action_space.seed(1)
    action = env.action_space.sample()
    actions.append(action)
print(actions[0] == actions[1])
# Outputs:
# [False False False False False False False False False False False False
# False False False False False]

System info

import envpool, numpy, sys
print(envpool.__version__, numpy.__version__, sys.version, sys.platform)

0.6.5 1.23.4 3.9.7 (default, Sep 16 2021, 13:09:58)
[GCC 7.5.0] linux

Checklist

• I have checked that there is no similar issue in the repo (required)
• I have read the documentation (required)
• I have provided a minimal working example to reproduce the bug (required)