Skip to content

[Deprecation] Prepare v0.15.0: postpone unimplemented default changes to v0.16 and bump the version - #4698

Merged
theap06 merged 5 commits into
pytorch:mainfrom
theap06:fix/v015-release-prep
Nov 7, 2026
Merged

theap06 merged 5 commits into
pytorch:mainfrom
theap06:fix/v015-release-prep

Conversation

@theap06

@theap06 theap06 commented Nov 6, 2026 •

Copy link
Copy Markdown
Collaborator

Description

Prepares main for the v0.15.0 cut.

  • Moves the promised v0.15 default changes, none of which are implemented, to v0.16: auto_register_policy_transforms, the AsyncBatchedCollector env backend and transport, the AsyncEnvPool exchange, and the unified checkpoint format for save_render_checkpoint and the Trainer. Current defaults are unchanged; warnings and docs now name v0.16.
  • Bumps the version to 0.15.0 in version.txt, pyproject.toml, td_script.sh and version_script.bat, and in the README CUDA install line.
  • Fixes docstrings that said auto_register_policy_transforms stays None "through v0.14".

Motivation and Context

Shipping v0.15.0 with warnings that say "this changes in v0.15" would break a published promise. CLAUDE.md section 12 asks for the target version to be named explicitly, so the warnings now name the release in which the change will actually happen.

  • I have raised an issue to propose this change (required for new features and bug fixes)

Types of changes

  • Documentation (update in the documentation)
  • Example (update in the folder of examples)

Checklist

  • I have read the CONTRIBUTION guide (required)
  • My change requires a change to the documentation.
  • I have updated the tests accordingly (required for a bug fix or a new feature).
  • I have updated the documentation accordingly.

Three FutureWarnings announced default changes for v0.15 that are not
implemented, and flipping them now is not safe:
- Collector and the multi-process collectors: auto_register_policy_transforms
  becomes True. The detection appends InitTracker for any nn.Module policy,
  which adds is_init entries to every batch.
- AsyncBatchedCollector: env_backend becomes 'multiprocessing' and
  transport becomes 'auto'.
- AsyncEnvPool with the multiprocessing backend: exchange becomes 'auto'.

The warnings, docstrings, docs, example and tutorial now name v0.16. The
current defaults are unchanged.
…0.16

save_render_checkpoint and the Trainer warned that their default output
would switch to the unified torchrl.checkpoint format in v0.15, and
neither switch is implemented. The warnings, docs and tests now name
v0.16; the current defaults are unchanged.
@pytorch-bot

pytorch-bot Bot commented Nov 6, 2026 •

Copy link
Copy Markdown

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4698

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 081b316 with merge base c5a5e3a (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Nov 6, 2026
@github-actions github-actions Bot added Documentation Improvements or additions to documentation CI Has to do with CI setup (e.g. wheels & builds, tests...) Examples Collectors tutorials/ Trainers Integrations/torch_geometric Integrations Deprecation labels Nov 6, 2026
@theap06
theap06 marked this pull request as ready for review November 7, 2026 07:11
@theap06
theap06 merged commit 003c5f4 into pytorch:main Nov 7, 2026
125 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CI Has to do with CI setup (e.g. wheels & builds, tests...) CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Collectors Deprecation Documentation Improvements or additions to documentation Examples Integrations/torch_geometric Integrations Trainers tutorials/

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant