[BugFix] PPOs with composite distribution #2791
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/2791
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 21 PendingAs of commit eadb9e1 with merge base 27a8ecc ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
louisfaury
commented
Feb 17, 2025
Comment on lines
+578
to
+581
| if is_tensor_collection(log_weight): | ||
| log_weight = _sum_td_features(log_weight) | ||
| log_weight = log_weight.view(adv_shape).unsqueeze(-1) | ||
|
|
There was a problem hiding this comment.
That is the main change for this method, which is also now consistent with type hints.
| @@ -987,8 +982,6 @@ def forward(self, tensordict: TensorDictBase) -> TensorDictBase: | |||
| # to different, unrelated trajectories, which is not standard. Still, it can give an idea of the weights' | |||
| # dispersion. | |||
| lw = log_weight.squeeze() | |||
| if not isinstance(lw, torch.Tensor): | |||
| lw = _sum_td_features(lw) | |||
| ess = (2 * lw.logsumexp(0) - (2 * lw).logsumexp(0)).exp() | |||
| batch = log_weight.shape[0] | |||
There was a problem hiding this comment.
The main error is two lines below; clamp was applied to the TensorDict log_weight before it is summed over the feature dimension.
vmoens
changed the title
vmoens
pushed a commit
that referenced
this pull request
Feb 17, 2025
Co-authored-by: Louis Faury <[email protected]> (cherry picked from commit edfa25d)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
I believe there is a bug in PPOs' implementation when both
prev_log_probandlog_probare TensorDicts.Motivation and Context
In the setting were both
prev_log_probandlog_probare TensorDicts, we were clampingprev_log_prob - log_probdirectly, instead of their sum over features.Types of changes
What types of changes does your code introduce? Remove all that do not apply:
Checklist
Go over all the following points, and put an
xin all the boxes that apply.If you are unsure about any of these, don't hesitate to ask. We are here to help!