Skip to content

Pull requests: pytorch/rl

New pull request New
108 Open 3,141 Closed
108 Open 3,141 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[Trainer] Implement GRPOTrainer and SOTA execution pipelines 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. Documentation Improvements or additions to documentation sota-implementations/ Trainers
#4140 opened Aug 21, 2026 by coder-jayp Contributor Loading…
7 of 8 tasks
4
[Deprecation] Make DreamerV3 defaults canonical Benchmarks rl/benchmark changes 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. Deprecation Documentation Improvements or additions to documentation Integrations/torch_geometric Integrations Modules Objectives sota-implementations/
#4134 opened Aug 20, 2026 by vmoens Collaborator • Draft
4
[Algorithm] Reproduce the pinned JAX DreamerV3 Walker preset Benchmarks rl/benchmark changes ci/optdeps Run the full tests-optdeps suite on this PR 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. Documentation Improvements or additions to documentation Integrations/torch_geometric Integrations Modules new algo New algorithm request or PR Objectives sota-implementations/
#4133 opened Aug 20, 2026 by vmoens Collaborator Loading…
1
[Performance] Speed up DreamerV3 RSSM rollout Benchmarks rl/benchmark changes ci/olddeps Run the tests-olddeps suite (oldest supported torch) on this PR 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. Documentation Improvements or additions to documentation Integrations/torch_geometric Integrations Modules Objectives Performance Performance issue or suggestion for improvement sota-implementations/
#4132 opened Aug 20, 2026 by vmoens Collaborator Loading…
2
[Performance] Improve geometric window sampling Benchmarks rl/benchmark changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Performance Performance issue or suggestion for improvement ReplayBuffers
#4131 opened Aug 19, 2026 by theap06 Contributor Loading…
3
[Feature] Add geometric trajectory window sampler Benchmarks rl/benchmark changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Examples Feature New feature ReplayBuffers
#4130 opened Aug 19, 2026 by vmoens Collaborator • Draft
1
[Feature] Opt-in percentile advantage scaling for PPO and A2C CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Feature New feature Objectives Trainers
#4127 opened Aug 18, 2026 by vmoens Collaborator • 3/3 Loading…
1
[Feature] Add PercentileValueNorm and refactor DreamerV3 retnorm onto it CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Integrations/torch_geometric Integrations Modules Objectives
#4126 opened Aug 18, 2026 by vmoens Collaborator • 2/3 Loading…
1
[BugFix] Apply DreamerV3 return normalization in the reparameterization branch BugFix CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Objectives
#4125 opened Aug 18, 2026 by vmoens Collaborator • 1/3 Loading…
1
Fix multiprocessing start method ValueError on Windows in tutorials CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. tutorials/
#4123 opened Aug 17, 2026 by coder-jayp Contributor Loading…
3 of 6 tasks
1
[Feature] Add FixedBatchedInference helper (AsyncEnvPool v2 item 3) CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Feature New feature tutorials/ user-facing User-facing changes - only include in major releases
#4110 opened Aug 15, 2026 by coder-jayp Contributor Loading…
5 of 6 tasks
17
[Feature] CraftGround (Minecraft) environment wrapper ci/optdeps Run the full tests-optdeps suite on this PR 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. Documentation Improvements or additions to documentation Environments/craftground Triggers craftground (Minecraft) environment tests only Environments Adds or modifies an environment wrapper Feature New feature user-facing User-facing changes - only include in major releases
#4107 opened Aug 14, 2026 by vmoens Collaborator Loading…
1
[Feature] WS1: Add stable component boundary Protocols for external-l… CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Feature New feature llm/ LLM-related PR, triggers LLM CI tests tutorials/ user-facing User-facing changes - only include in major releases
#4103 opened Aug 13, 2026 by coder-jayp Contributor Loading…
7 tasks done
1
12
[Feature] Wire value transforms into actor-critic losses Benchmarks rl/benchmark changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Feature New feature Integrations/torch_geometric Integrations Modules Objectives Trainers user-facing User-facing changes - only include in major releases
#4100 opened Aug 13, 2026 by vmoens Collaborator Loading…
1
Bump torch from 2.12.0 to 2.13.0 in /sota-implementations/expert-iteration CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Dependencies Pull requests that update a dependency file python Pull requests that update python code sota-implementations/
#4037 opened Jul 17, 2026 by dependabot Bot Loading…
5
[Performance] Avoid parent env construction in ParallelEnv startup Benchmarks rl/benchmark changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Performance Performance issue or suggestion for improvement sota-implementations/
#4029 opened Jul 16, 2026 by vmoens Collaborator • Draft
4
[BugFix] Stage collector weight sync through CPU when device sharing … BugFix CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Collectors Integrations/torch_geometric Integrations WeightUpdate
#4027 opened Jul 16, 2026 by theap06 Contributor Loading…
1
1
[Test] Add off-policy data-parallel learning suite CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. sota-implementations/ Tests Incomplete or broken unit tests Trainers
#4026 opened Jul 16, 2026 by vmoens Collaborator • Draft
2
[Feature] Add reliable VLA-GRPO LIBERO recipe 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. Feature New feature llm/ LLM-related PR, triggers LLM CI tests Objectives sota-implementations/ Transforms
#4024 opened Jul 16, 2026 by vmoens Collaborator • Draft
2
[Feature] Add flat replay writes to collectors Benchmarks rl/benchmark changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Collectors Documentation Improvements or additions to documentation Feature New feature Integrations/torch_geometric Integrations Trainers
#4020 opened Jul 16, 2026 by vmoens Collaborator Loading…
1
[BugFix] Bound direct off-policy replay collection BugFix CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Collectors Integrations/torch_geometric Integrations ReplayBuffers Trainers
#4018 opened Jul 16, 2026 by vmoens Collaborator Loading…
1
[LLM] WS4 Post-Training Observability: Implement PostTrainingLogger CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation llm/ LLM-related PR, triggers LLM CI tests Objectives Record sota-implementations/
#3950 opened Jul 7, 2026 by coder-jayp Contributor Loading…
7 of 8 tasks
1
13
[Feature] Add Learner primitive (LocalLearner, FSDP2Learner) CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Feature New feature Trainers user-facing User-facing changes - only include in major releases
#3926 opened Jul 2, 2026 by theap06 Contributor Loading…
2
[Algorithm] Add RLHF reward-model training recipe 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. Documentation Improvements or additions to documentation llm/ LLM-related PR, triggers LLM CI tests new algo New algorithm request or PR Objectives sota-implementations/ user-facing User-facing changes - only include in major releases
#3923 opened Jun 30, 2026 by theap06 Contributor Loading…
15
[Feature] Add RewardModelLoss objective for RLHF reward-model training CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. Documentation Improvements or additions to documentation Feature New feature llm/ LLM-related PR, triggers LLM CI tests Objectives user-facing User-facing changes - only include in major releases
#3922 opened Jun 30, 2026 by theap06 Contributor Loading…
7
ProTip! Follow long discussions with comments:>50.