Skip to content

Pull requests: PrimeIntellect-ai/prime-rl

New pull request New
138 Open 2,630 Closed
138 Open 2,630 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

fix(nixl): replay strided views on inference
#3113 opened Jul 22, 2026 by S1ro1 Collaborator Loading…
feat(wandb): log full RLConfig to W&B instead of per-component subconfig
#3110 opened Jul 22, 2026 by samsja Member • Draft
2
feat: chunked per-layer optimizer step with stream-overlapped CPU offload
#3109 opened Jul 22, 2026 by samsja Member Loading…
5
fix(orchestrator): gate group-relative algorithms to a single trainable agent
#3108 opened Jul 22, 2026 by hallerite Member Loading…
3
fix(orchestrator): drain after final train batch
#3107 opened Jul 22, 2026 by jooooonas • Draft
chore: single global router in front of engines for all deployments
#3105 opened Jul 22, 2026 by mikasenghaas Member Loading…
fix: replace wandb_gql import with execute_graphql for wandb>=0.27.1 compat
#3086 opened Jul 19, 2026 by TM23-sanji Loading…
4
fix(sft): tolerate trace-shape tool calls; opt-in skip for unrenderable samples
#3081 opened Jul 18, 2026 by mikasenghaas Member Loading…
feat: always-on UUID log directories with symlink
#3074 opened Jul 17, 2026 by samsja Member Loading…
3
feat: DSA conversion (dense-MLA -> DeepSeek Sparse Attention), with a Kimi K2 example
#3066 opened Jul 17, 2026 by faresobeid Contributor • Draft
feat(models): add Kimi K2 + K2.7 (text-only) support
#3065 opened Jul 17, 2026 by faresobeid Contributor • Draft
feat: dynamo prime-rl integration
#3062 opened Jul 16, 2026 by biswapanda Loading…
21
chore: bump transformers to 5.14.1
#3055 opened Jul 16, 2026 by hallerite Member Loading…
3
fix(orchestrator): hold batch ship on the trainer's published version
#3050 opened Jul 16, 2026 by mikasenghaas Member Loading…
21
fix(orchestrator): coordinate final NCCL shutdown
#3047 opened Jul 15, 2026 by kalomaze Contributor • Draft
fix(trainer): stage CPU-offloaded weights through CUDA in NCCL broadcast
#3045 opened Jul 15, 2026 by mikasenghaas Member • Draft
5
fix(trainer): require complete multi-run checkpoint publication
#3038 opened Jul 15, 2026 by peter941221 Contributor Loading…
1
fix(orchestrator): pin eval epochs to one policy version
#3034 opened Jul 15, 2026 by vineetjain96 Loading…
fix(orchestrator): align train metrics with trainer cohort
#3033 opened Jul 15, 2026 by vineetjain96 Loading…
refactor: share training batch production
#3027 opened Jul 14, 2026 by tim0120 Contributor • Draft
decouple trainer step completion from weight broadcast
#3024 opened Jul 14, 2026 by tim0120 Contributor • Draft
[ENG-4450] feature: prime conversion entrypoint
#3021 opened Jul 13, 2026 by eexwhyzee Collaborator Loading…
7
feat(orchestrator): use the renderer client for evals
#3017 opened Jul 13, 2026 by faresobeid Contributor • Draft
fix(orchestrator): unify off-policy lag across sink queue
#3009 opened Jul 13, 2026 by Abhishek21g Loading…
2
feat(orchestrator): expose target_lag config
#3008 opened Jul 13, 2026 by Abhishek21g Loading…
2
ProTip! Follow long discussions with comments:>50.