-
Notifications
You must be signed in to change notification settings - Fork 5.8k
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
[RLlib] 2 fixes for 2.10 release: 1) DDP + GPU fix; 2) py3.8 fix for uniting two dicts #44000
[RLlib] 2 fixes for 2.10 release: 1) DDP + GPU fix; 2) py3.8 fix for uniting two dicts #44000
Conversation
Signed-off-by: sven1977 <[email protected]>
@@ -871,7 +871,7 @@ def step(self) -> ResultDict: | |||
results = self._compile_iteration_results( | |||
episodes_this_iter=episodes_this_iter, | |||
step_ctx=train_iter_ctx, | |||
iteration_results=train_results | eval_results, | |||
iteration_results={**train_results, **eval_results}, |
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.
py3.8 does not support dict | dict
.
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 py3.8 deprecated by the next Ray release version?
@@ -183,6 +183,13 @@ def _save_module_metadata(self, *args, **kwargs): | |||
def _module_metadata(self, *args, **kwargs): | |||
return self.unwrapped()._module_metadata(*args, **kwargs) | |||
|
|||
# TODO (sven): Figure out a better way to avoid having to method-spam this wrapper |
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.
We need to find a better solution for this, but this fixes PPO + multi-GPU on the new stack for now.
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
@@ -871,7 +871,7 @@ def step(self) -> ResultDict: | |||
results = self._compile_iteration_results( | |||
episodes_this_iter=episodes_this_iter, | |||
step_ctx=train_iter_ctx, | |||
iteration_results=train_results | eval_results, | |||
iteration_results={**train_results, **eval_results}, |
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 py3.8 deprecated by the next Ray release version?
# class, whenever we add a new API to any wrapped RLModule here. We could try | ||
# auto generating the wrapper methods, but this will bring its own challenge | ||
# (e.g. recursive calls due to __getattr__ checks, etc..). | ||
def _compute_values(self, *args, **kwargs): |
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.
Btw: Can't we also compile this method?
Cherry pick #44000 Signed-off-by: sven1977 <[email protected]>
2 fixes for 2.10 release: 1) DDP + GPU fix; 2) py3.8 fix for uniting two dicts
Why are these changes needed?
Related issue number
Checks
git commit -s
) in this PR.scripts/format.sh
to lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/
under thecorresponding
.rst
file.