Unable to run l2rpn_neurips_2020_track1_small using Rllib integration code #196
Comments
rrhossain
changed the title
|
Hello, I will have a look at the first error. It seems that the upper / lower bound of the observation space are too tight. I will try to reproduce it and fix it asap. With the provided error, it does not seem to be an issue though (i mean: the observation given appears to be in the observation space, so this might - i say might- be a ray / rllib or a gym error and not a grid2op error) For the second one, I am aware that multiprocessing does not work well on macos / windows due to internal working of this package.
There is for now little i can do about it unfortunately. It works fine on linux based machine and on macos with python 3.6 or 3.7 i believe. They "broke" it starting from python 3.8 on macOs if i recall correctly. There might be a good chance that this code works if you run it outside a jupyter notebook though (running multiprocessing in jupyter notebook on MacOs and Windows never works pretty much, see eg https://stackoverflow.com/questions/23641475/multiprocessing-working-in-python-but-not-in-ipython/23641560#23641560 for a more detailed explanation) |
|
Hi, The issue with the "obs not in obs space" (or something like that) is indeed due to :
You can fix them with: class MyEnv(gym.Env):
def __init__(self, env_config):
import grid2op
from grid2op.gym_compat import GymEnv
from grid2op.gym_compat import ScalerAttrConverter, ContinuousToDiscreteConverter, MultiToTupleConverter
from lightsim2grid import LightSimBackend
# 1. create the grid2op environment
if not "env_name" in env_config:
raise RuntimeError("The configuration for RLLIB should provide the env name")
nm_env = env_config["env_name"]
del env_config["env_name"]
self.env_glop = grid2op.make(nm_env, backend=LightSimBackend(), **env_config)
self.env_glop.deactivate_forecast()
# 2. create the gym environment
self.env_gym = GymEnv(self.env_glop)
obs_gym = self.env_gym.reset()
# 3. (optional) customize it (see section above for more information)
## customize action space
self.env_gym.action_space = self.env_gym.action_space.ignore_attr("set_bus").ignore_attr("set_line_status")
self.env_gym.action_space = self.env_gym.action_space.reencode_space("redispatch",
ContinuousToDiscreteConverter(nb_bins=11)
)
self.env_gym.action_space = self.env_gym.action_space.reencode_space("change_bus", MultiToTupleConverter())
self.env_gym.action_space = self.env_gym.action_space.reencode_space("change_line_status",
MultiToTupleConverter())
self.env_gym.action_space = self.env_gym.action_space.reencode_space("redispatch", MultiToTupleConverter())
## customize observation space
ob_space = self.env_gym.observation_space
ob_space = ob_space.keep_only_attr(["rho", "gen_p", "load_p", "topo_vect", "actual_dispatch"])
if True:
ob_space = ob_space.reencode_space("actual_dispatch",
ScalerAttrConverter(substract=0.,
divide=self.env_glop.gen_pmax
)
)
ob_space = ob_space.reencode_space("gen_p",
ScalerAttrConverter(substract=0.,
divide=self.env_glop.gen_pmax
)
)
ob_space = ob_space.reencode_space("load_p",
ScalerAttrConverter(substract=obs_gym["load_p"],
divide=0.5 * obs_gym["load_p"]
)
)
self.env_gym.observation_space = ob_space
# 4. specific to rllib
self.action_space = self.env_gym.action_space
self.observation_space = self.env_gym.observation_space
self.step_count = 0
self.observation_space["gen_p"].low[:] = -np.inf
self.observation_space["gen_p"].high[:] = np.inf
self.observation_space["load_p"].low[:] = -np.inf
self.observation_space["load_p"].high[:] = np.inf
def reset(self):
obs = self.env_gym.reset()
self.step_count = 0
return obs
def step(self, action):
self.step_count += 1
obs, reward, done, info = self.env_gym.step(action)
return obs, reward, done, infoA fix (partial for the first issue) will come soon. The "complete" fix for the issue is to put |
|
Code is working on linux now |
Environment
1.5.1Bug description
The code is working for l2rpn-sandbox but it is producing error for l2rpn-neurips-track-1-small environment.
Code snippet
The output of the code snippet above (in colab):
The output of the code snippet above (in local machine)
The text was updated successfully, but these errors were encountered: