-
Notifications
You must be signed in to change notification settings - Fork 5.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Revert "Revert "Revert "Revert "[RLlib] Upgrade to gymnasium 1.0.0 (ale_py 0.10.1, mujoco 3.2.4, pettingzoo 1.24.3 supersuit 3.9.3)."""" #48443
Conversation
…le_py 0.…" This reverts commit 0c58e87.
will leave this to @can-anyscale and other people. let me know if you need explicit approval on the we might be breaking a record of the number of "revert"'s in the title.. maybe consider breaking this PR into smaller ones so that it is easier to make progress(though I am not sure what is the best way to break it)? maybe we can accept using |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
all power to @can-anyscale
…rt-48435-revert-48308-revert-48297-revert-45328-upgrade_gymnasium_to_1_0_0a1
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM. Huge change now. Maybe I overread it, but could you change the environment also in the BC benchmark with RLUnplugged?
@@ -259,7 +263,7 @@ def _sample_timesteps( | |||
to_env = { | |||
Columns.ACTIONS: [ | |||
{ | |||
aid: self.env.get_action_space(aid).sample() | |||
aid: self.env.unwrapped.get_action_space(aid).sample() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So this is now the new thing: always unwrap?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is only RLlib-internally b/c the vector env is wrapped to return info-lists (rather than dicts).
I think we still have to unify the way the user sees these envs in their callbacks: Always as-is (vector wrapper) or always as unwrapped (pure gym.vector.Env). Both approaches have advantages and disadvantages.
|
||
ts += self.num_envs | ||
observations, rewards, terminateds, truncateds, infos = results | ||
observations, actions = unbatch(observations), unbatch(actions) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Isn't there a gymnasium wrapper that does this?
"rllib-single-agent-env-v0", | ||
num_envs=self.config.num_envs_per_env_runner, | ||
asynchronous=self.config.remote_worker_envs, | ||
vectorization_mode=( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Very nice! I remember in my problems with the update nto gymnasium==1.0.0 this was one reason of failure. It needs now this mode.
…rt-48435-revert-48308-revert-48297-revert-45328-upgrade_gymnasium_to_1_0_0a1
Hey @aslonnie , could we give this a go? Release tests are passing properly now: |
…le_py 0.10.1, mujoco 3.2.4, pettingzoo 1.24.3 supersuit 3.9.3)."""" (ray-project#48443)
…le_py 0.10.1, mujoco 3.2.4, pettingzoo 1.24.3 supersuit 3.9.3)."""" (ray-project#48443) Signed-off-by: JP-sDEV <jon.pablo80@gmail.com>
…le_py 0.10.1, mujoco 3.2.4, pettingzoo 1.24.3 supersuit 3.9.3)."""" (ray-project#48443) Signed-off-by: mohitjain2504 <mohit.jain@dream11.com>
Reverts #48435