GameServerO / MLPY /Lib /site-packages /gym /utils /passive_env_checker.py
Kano001's picture
Upload 919 files
375a1cf verified
raw
history blame
15 kB
"""A set of functions for passively checking environment implementations."""
import inspect
from functools import partial
from typing import Callable
import numpy as np
from gym import Space, error, logger, spaces
def _check_box_observation_space(observation_space: spaces.Box):
"""Checks that a :class:`Box` observation space is defined in a sensible way.
Args:
observation_space: A box observation space
"""
# Check if the box is an image
if len(observation_space.shape) == 3:
if observation_space.dtype != np.uint8:
logger.warn(
f"It seems a Box observation space is an image but the `dtype` is not `np.uint8`, actual type: {observation_space.dtype}. "
"If the Box observation space is not an image, we recommend flattening the observation to have only a 1D vector."
)
if np.any(observation_space.low != 0) or np.any(observation_space.high != 255):
logger.warn(
"It seems a Box observation space is an image but the upper and lower bounds are not in [0, 255]. "
"Generally, CNN policies assume observations are within that range, so you may encounter an issue if the observation values are not."
)
if len(observation_space.shape) not in [1, 3]:
logger.warn(
"A Box observation space has an unconventional shape (neither an image, nor a 1D vector). "
"We recommend flattening the observation to have only a 1D vector or use a custom policy to properly process the data. "
f"Actual observation shape: {observation_space.shape}"
)
assert (
observation_space.low.shape == observation_space.shape
), f"The Box observation space shape and low shape have different shapes, low shape: {observation_space.low.shape}, box shape: {observation_space.shape}"
assert (
observation_space.high.shape == observation_space.shape
), f"The Box observation space shape and high shape have have different shapes, high shape: {observation_space.high.shape}, box shape: {observation_space.shape}"
if np.any(observation_space.low == observation_space.high):
logger.warn("A Box observation space maximum and minimum values are equal.")
elif np.any(observation_space.high < observation_space.low):
logger.warn("A Box observation space low value is greater than a high value.")
def _check_box_action_space(action_space: spaces.Box):
"""Checks that a :class:`Box` action space is defined in a sensible way.
Args:
action_space: A box action space
"""
assert (
action_space.low.shape == action_space.shape
), f"The Box action space shape and low shape have have different shapes, low shape: {action_space.low.shape}, box shape: {action_space.shape}"
assert (
action_space.high.shape == action_space.shape
), f"The Box action space shape and high shape have different shapes, high shape: {action_space.high.shape}, box shape: {action_space.shape}"
if np.any(action_space.low == action_space.high):
logger.warn("A Box action space maximum and minimum values are equal.")
elif np.any(action_space.high < action_space.low):
logger.warn("A Box action space low value is greater than a high value.")
def check_space(
space: Space, space_type: str, check_box_space_fn: Callable[[spaces.Box], None]
):
"""A passive check of the environment action space that should not affect the environment."""
if not isinstance(space, spaces.Space):
raise AssertionError(
f"{space_type} space does not inherit from `gym.spaces.Space`, actual type: {type(space)}"
)
elif isinstance(space, spaces.Box):
check_box_space_fn(space)
elif isinstance(space, spaces.Discrete):
assert (
0 < space.n
), f"Discrete {space_type} space's number of elements must be positive, actual number of elements: {space.n}"
assert (
space.shape == ()
), f"Discrete {space_type} space's shape should be empty, actual shape: {space.shape}"
elif isinstance(space, spaces.MultiDiscrete):
assert (
space.shape == space.nvec.shape
), f"Multi-discrete {space_type} space's shape must be equal to the nvec shape, space shape: {space.shape}, nvec shape: {space.nvec.shape}"
assert np.all(
0 < space.nvec
), f"Multi-discrete {space_type} space's all nvec elements must be greater than 0, actual nvec: {space.nvec}"
elif isinstance(space, spaces.MultiBinary):
assert np.all(
0 < np.asarray(space.shape)
), f"Multi-binary {space_type} space's all shape elements must be greater than 0, actual shape: {space.shape}"
elif isinstance(space, spaces.Tuple):
assert 0 < len(
space.spaces
), f"An empty Tuple {space_type} space is not allowed."
for subspace in space.spaces:
check_space(subspace, space_type, check_box_space_fn)
elif isinstance(space, spaces.Dict):
assert 0 < len(
space.spaces.keys()
), f"An empty Dict {space_type} space is not allowed."
for subspace in space.values():
check_space(subspace, space_type, check_box_space_fn)
check_observation_space = partial(
check_space,
space_type="observation",
check_box_space_fn=_check_box_observation_space,
)
check_action_space = partial(
check_space, space_type="action", check_box_space_fn=_check_box_action_space
)
def check_obs(obs, observation_space: spaces.Space, method_name: str):
"""Check that the observation returned by the environment correspond to the declared one.
Args:
obs: The observation to check
observation_space: The observation space of the observation
method_name: The method name that generated the observation
"""
pre = f"The obs returned by the `{method_name}()` method"
if isinstance(observation_space, spaces.Discrete):
if not isinstance(obs, (np.int64, int)):
logger.warn(f"{pre} should be an int or np.int64, actual type: {type(obs)}")
elif isinstance(observation_space, spaces.Box):
if observation_space.shape != ():
if not isinstance(obs, np.ndarray):
logger.warn(
f"{pre} was expecting a numpy array, actual type: {type(obs)}"
)
elif obs.dtype != observation_space.dtype:
logger.warn(
f"{pre} was expecting numpy array dtype to be {observation_space.dtype}, actual type: {obs.dtype}"
)
elif isinstance(observation_space, (spaces.MultiBinary, spaces.MultiDiscrete)):
if not isinstance(obs, np.ndarray):
logger.warn(f"{pre} was expecting a numpy array, actual type: {type(obs)}")
elif isinstance(observation_space, spaces.Tuple):
if not isinstance(obs, tuple):
logger.warn(f"{pre} was expecting a tuple, actual type: {type(obs)}")
assert len(obs) == len(
observation_space.spaces
), f"{pre} length is not same as the observation space length, obs length: {len(obs)}, space length: {len(observation_space.spaces)}"
for sub_obs, sub_space in zip(obs, observation_space.spaces):
check_obs(sub_obs, sub_space, method_name)
elif isinstance(observation_space, spaces.Dict):
assert isinstance(obs, dict), f"{pre} must be a dict, actual type: {type(obs)}"
assert (
obs.keys() == observation_space.spaces.keys()
), f"{pre} observation keys is not same as the observation space keys, obs keys: {list(obs.keys())}, space keys: {list(observation_space.spaces.keys())}"
for space_key in observation_space.spaces.keys():
check_obs(obs[space_key], observation_space[space_key], method_name)
try:
if obs not in observation_space:
logger.warn(f"{pre} is not within the observation space.")
except Exception as e:
logger.warn(f"{pre} is not within the observation space with exception: {e}")
def env_reset_passive_checker(env, **kwargs):
"""A passive check of the `Env.reset` function investigating the returning reset information and returning the data unchanged."""
signature = inspect.signature(env.reset)
if "seed" not in signature.parameters and "kwargs" not in signature.parameters:
logger.warn(
"Future gym versions will require that `Env.reset` can be passed a `seed` instead of using `Env.seed` for resetting the environment random number generator."
)
else:
seed_param = signature.parameters.get("seed")
# Check the default value is None
if seed_param is not None and seed_param.default is not None:
logger.warn(
"The default seed argument in `Env.reset` should be `None`, otherwise the environment will by default always be deterministic. "
f"Actual default: {seed_param}"
)
if "options" not in signature.parameters and "kwargs" not in signature.parameters:
logger.warn(
"Future gym versions will require that `Env.reset` can be passed `options` to allow the environment initialisation to be passed additional information."
)
# Checks the result of env.reset with kwargs
result = env.reset(**kwargs)
if not isinstance(result, tuple):
logger.warn(
f"The result returned by `env.reset()` was not a tuple of the form `(obs, info)`, where `obs` is a observation and `info` is a dictionary containing additional information. Actual type: `{type(result)}`"
)
elif len(result) != 2:
logger.warn(
"The result returned by `env.reset()` should be `(obs, info)` by default, , where `obs` is a observation and `info` is a dictionary containing additional information."
)
else:
obs, info = result
check_obs(obs, env.observation_space, "reset")
assert isinstance(
info, dict
), f"The second element returned by `env.reset()` was not a dictionary, actual type: {type(info)}"
return result
def env_step_passive_checker(env, action):
"""A passive check for the environment step, investigating the returning data then returning the data unchanged."""
# We don't check the action as for some environments then out-of-bounds values can be given
result = env.step(action)
assert isinstance(
result, tuple
), f"Expects step result to be a tuple, actual type: {type(result)}"
if len(result) == 4:
logger.deprecation(
"Core environment is written in old step API which returns one bool instead of two. "
"It is recommended to rewrite the environment with new step API. "
)
obs, reward, done, info = result
if not isinstance(done, (bool, np.bool8)):
logger.warn(
f"Expects `done` signal to be a boolean, actual type: {type(done)}"
)
elif len(result) == 5:
obs, reward, terminated, truncated, info = result
# np.bool is actual python bool not np boolean type, therefore bool_ or bool8
if not isinstance(terminated, (bool, np.bool8)):
logger.warn(
f"Expects `terminated` signal to be a boolean, actual type: {type(terminated)}"
)
if not isinstance(truncated, (bool, np.bool8)):
logger.warn(
f"Expects `truncated` signal to be a boolean, actual type: {type(truncated)}"
)
else:
raise error.Error(
f"Expected `Env.step` to return a four or five element tuple, actual number of elements returned: {len(result)}."
)
check_obs(obs, env.observation_space, "step")
if not (
np.issubdtype(type(reward), np.integer)
or np.issubdtype(type(reward), np.floating)
):
logger.warn(
f"The reward returned by `step()` must be a float, int, np.integer or np.floating, actual type: {type(reward)}"
)
else:
if np.isnan(reward):
logger.warn("The reward is a NaN value.")
if np.isinf(reward):
logger.warn("The reward is an inf value.")
assert isinstance(
info, dict
), f"The `info` returned by `step()` must be a python dictionary, actual type: {type(info)}"
return result
def env_render_passive_checker(env, *args, **kwargs):
"""A passive check of the `Env.render` that the declared render modes/fps in the metadata of the environment is declared."""
render_modes = env.metadata.get("render_modes")
if render_modes is None:
logger.warn(
"No render modes was declared in the environment (env.metadata['render_modes'] is None or not defined), you may have trouble when calling `.render()`."
)
else:
if not isinstance(render_modes, (list, tuple)):
logger.warn(
f"Expects the render_modes to be a sequence (i.e. list, tuple), actual type: {type(render_modes)}"
)
elif not all(isinstance(mode, str) for mode in render_modes):
logger.warn(
f"Expects all render modes to be strings, actual types: {[type(mode) for mode in render_modes]}"
)
render_fps = env.metadata.get("render_fps")
# We only require `render_fps` if rendering is actually implemented
if len(render_modes) > 0:
if render_fps is None:
logger.warn(
"No render fps was declared in the environment (env.metadata['render_fps'] is None or not defined), rendering may occur at inconsistent fps."
)
else:
if not (
np.issubdtype(type(render_fps), np.integer)
or np.issubdtype(type(render_fps), np.floating)
):
logger.warn(
f"Expects the `env.metadata['render_fps']` to be an integer or a float, actual type: {type(render_fps)}"
)
else:
assert (
render_fps > 0
), f"Expects the `env.metadata['render_fps']` to be greater than zero, actual value: {render_fps}"
# env.render is now an attribute with default None
if len(render_modes) == 0:
assert (
env.render_mode is None
), f"With no render_modes, expects the Env.render_mode to be None, actual value: {env.render_mode}"
else:
assert env.render_mode is None or env.render_mode in render_modes, (
"The environment was initialized successfully however with an unsupported render mode. "
f"Render mode: {env.render_mode}, modes: {render_modes}"
)
result = env.render(*args, **kwargs)
# TODO: Check that the result is correct
return result