mirror of
https://github.com/saymrwulf/stable-baselines3.git
synced 2026-11-18 09:54:10 +00:00
Fix env checker single-step-env edge case (#1524)
* Fix env checker single-step-env edge case Before this change, env checker failed to `reset()` the tested environment before calling `step()` when checking for `Inf` / `NaN`. This could cause environments which happened to have only one `step()` available before the episode was terminated to fail. This is now fixed. * Code review fixes #1 As suggested by Antonin Raffin <antonin.raffin@ensta.org>.
This commit is contained in:
parent
1bfb55df0e
commit
e76316341d
4 changed files with 51 additions and 3 deletions
|
|
@ -3,7 +3,7 @@
|
||||||
Changelog
|
Changelog
|
||||||
==========
|
==========
|
||||||
|
|
||||||
Release 2.0.0a10 (WIP)
|
Release 2.0.0a11 (WIP)
|
||||||
--------------------------
|
--------------------------
|
||||||
|
|
||||||
**Gymnasium support**
|
**Gymnasium support**
|
||||||
|
|
@ -39,6 +39,8 @@ Bug Fixes:
|
||||||
- Fixed ``VecExtractDictObs`` does not handle terminal observation (@WeberSamuel)
|
- Fixed ``VecExtractDictObs`` does not handle terminal observation (@WeberSamuel)
|
||||||
- Set NumPy version to ``>=1.20`` due to use of ``numpy.typing`` (@troiganto)
|
- Set NumPy version to ``>=1.20`` due to use of ``numpy.typing`` (@troiganto)
|
||||||
- Fixed loading DQN changes ``target_update_interval`` (@tobirohrer)
|
- Fixed loading DQN changes ``target_update_interval`` (@tobirohrer)
|
||||||
|
- Fixed env checker to properly reset the env before calling ``step()`` when checking
|
||||||
|
for ``Inf`` and ``NaN`` (@lutogniew)
|
||||||
|
|
||||||
Deprecations:
|
Deprecations:
|
||||||
^^^^^^^^^^^^^
|
^^^^^^^^^^^^^
|
||||||
|
|
@ -1346,3 +1348,4 @@ And all the contributors:
|
||||||
@Melanol @qgallouedec @francescoluciano @jlp-ue @burakdmb @timothe-chaumont @honglu2875
|
@Melanol @qgallouedec @francescoluciano @jlp-ue @burakdmb @timothe-chaumont @honglu2875
|
||||||
@anand-bala @hughperkins @sidney-tio @AlexPasqua @dominicgkerr @Akhilez @Rocamonde @tobirohrer @ZikangXiong
|
@anand-bala @hughperkins @sidney-tio @AlexPasqua @dominicgkerr @Akhilez @Rocamonde @tobirohrer @ZikangXiong
|
||||||
@DavyMorgan @luizapozzobon @Bonifatius94 @theSquaredError @harveybellini @DavyMorgan @FieteO @jonasreiher @npit @WeberSamuel @troiganto
|
@DavyMorgan @luizapozzobon @Bonifatius94 @theSquaredError @harveybellini @DavyMorgan @FieteO @jonasreiher @npit @WeberSamuel @troiganto
|
||||||
|
@lutogniew
|
||||||
|
|
|
||||||
|
|
@ -110,6 +110,7 @@ def _check_unsupported_spaces(env: gym.Env, observation_space: spaces.Space, act
|
||||||
def _check_nan(env: gym.Env) -> None:
|
def _check_nan(env: gym.Env) -> None:
|
||||||
"""Check for Inf and NaN using the VecWrapper."""
|
"""Check for Inf and NaN using the VecWrapper."""
|
||||||
vec_env = VecCheckNan(DummyVecEnv([lambda: env]))
|
vec_env = VecCheckNan(DummyVecEnv([lambda: env]))
|
||||||
|
vec_env.reset()
|
||||||
for _ in range(10):
|
for _ in range(10):
|
||||||
action = np.array([env.action_space.sample()])
|
action = np.array([env.action_space.sample()])
|
||||||
_, _, _, _ = vec_env.step(action)
|
_, _, _, _ = vec_env.step(action)
|
||||||
|
|
|
||||||
|
|
@ -1 +1 @@
|
||||||
2.0.0a10
|
2.0.0a11
|
||||||
|
|
|
||||||
|
|
@ -1,4 +1,4 @@
|
||||||
from typing import Dict, Optional
|
from typing import Any, Dict, Optional, Tuple
|
||||||
|
|
||||||
import gymnasium as gym
|
import gymnasium as gym
|
||||||
import numpy as np
|
import numpy as np
|
||||||
|
|
@ -112,3 +112,47 @@ def test_check_env_detailed_error(obs_tuple, method):
|
||||||
test_env = TestEnv()
|
test_env = TestEnv()
|
||||||
with pytest.raises(AssertionError, match=error_message):
|
with pytest.raises(AssertionError, match=error_message):
|
||||||
check_env(env=test_env)
|
check_env(env=test_env)
|
||||||
|
|
||||||
|
|
||||||
|
class LimitedStepsTestEnv(gym.Env):
|
||||||
|
action_space = spaces.Discrete(n=2)
|
||||||
|
observation_space = spaces.Discrete(n=2)
|
||||||
|
|
||||||
|
def __init__(self, steps_before_termination: int = 1):
|
||||||
|
super().__init__()
|
||||||
|
|
||||||
|
assert steps_before_termination >= 1
|
||||||
|
self._steps_before_termination = steps_before_termination
|
||||||
|
|
||||||
|
self._steps_called = 0
|
||||||
|
self._terminated = False
|
||||||
|
|
||||||
|
def reset(self, *, seed: Optional[int] = None, options: Optional[Dict] = None) -> Tuple[int, Dict]:
|
||||||
|
super().reset(seed=seed)
|
||||||
|
|
||||||
|
self._steps_called = 0
|
||||||
|
self._terminated = False
|
||||||
|
|
||||||
|
return 0, {}
|
||||||
|
|
||||||
|
def step(self, action: np.ndarray) -> Tuple[int, float, bool, bool, Dict[str, Any]]:
|
||||||
|
self._steps_called += 1
|
||||||
|
|
||||||
|
assert not self._terminated
|
||||||
|
|
||||||
|
observation = 0
|
||||||
|
reward = 0.0
|
||||||
|
self._terminated = self._steps_called >= self._steps_before_termination
|
||||||
|
truncated = False
|
||||||
|
|
||||||
|
return observation, reward, self._terminated, truncated, {}
|
||||||
|
|
||||||
|
def render(self) -> None:
|
||||||
|
pass
|
||||||
|
|
||||||
|
|
||||||
|
def test_check_env_single_step_env():
|
||||||
|
test_env = LimitedStepsTestEnv(steps_before_termination=1)
|
||||||
|
|
||||||
|
# This should not throw
|
||||||
|
check_env(env=test_env, warn=True)
|
||||||
|
|
|
||||||
Loading…
Reference in a new issue