Ignore unused parameters when using DDP + PPG #1117
Merged
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Outdated suggestions cannot be applied.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.
Motivation
To make DDP + PPG work, even with #1116, DDP will still panic about unused parameters. Note that there are 2 isolated problems:
#1116 aims to fix No.1, and we need to instruct DDP to proactively prune the unused parameters. Though this will adds a little overhead but it is necessary for PPG.
Solution
Add a globally configurable parameters to find and ignore unused parameters for DDP. By default it is disabled but for PPG it is explicitly enabled.
Testing
Tested together with #1116 on ac/ppo/ppg cart pole to make sure all of them trains correct. PPG procgen is trained and reported in #1099.