Skip to content

Pull requests: PrimeIntellect-ai/prime-rl

New pull request New
77 Open 1,706 Closed
77 Open 1,706 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

feat: add NemotronH (Nemotron-3-Super-120B-A12B) model support
#2046 opened Mar 19, 2026 by samsja Loading…
6 of 7 tasks
2
feat: skip master weight gathering flag
#2045 opened Mar 18, 2026 by S1ro1 Loading…
feat: collect and log per-engine inference server metrics to W&B
#2043 opened Mar 18, 2026 by mikasenghaas • Draft
2
Add token usage reporting for billing
#2042 opened Mar 18, 2026 by shayonj Loading…
1 task
6
vLLM monkey-patches for dcp support on tp > node gpu count
#2041 opened Mar 18, 2026 by JannikSt Loading…
20
fix: compute GR3 baseline from length-corrected rewards
#2040 opened Mar 18, 2026 by rasdani • Draft
feat: integrate FP8 weight transfer format
#2038 opened Mar 17, 2026 by S1ro1 Loading…
2
feat: add sample-level loss normalization for SFT
#2032 opened Mar 16, 2026 by hallerite Loading…
9 tasks done
7
Fix config validation, engine resume, and CHANGELOG
#2031 opened Mar 16, 2026 by cursor bot • Draft
feat: disaggregated prefill-decode inference with multi-replica support
#2030 opened Mar 16, 2026 by samsja Loading…
4 tasks
37
WIP chore: refactor docs into structured sections
#2022 opened Mar 12, 2026 by samsja Loading…
3 tasks
fix vlm missing arg
#2021 opened Mar 12, 2026 by leopold-tzafon Loading…
7
fix: reconcile VLM image tokens with pixel_values after prompt truncation
#2013 opened Mar 10, 2026 by cdreetz Loading…
3 tasks done
3
perf: reduce orchestrator scheduler refill from O(k*n) to O(k*c)
#2011 opened Mar 10, 2026 by juvi21 Loading…
GLM5 Branch
#2010 opened Mar 10, 2026 by S1ro1 Loading…
1 of 2 tasks
17
feat: add Qwen 3.5 MoE model support
#2007 opened Mar 10, 2026 by samsja Loading…
3 tasks
5
perf: pipeline orchestrator postprocessing with next step's rollout generation to reduce inference idle time [wip]
#2002 opened Mar 9, 2026 by juvi21 Loading…
13
fix: support VLM models in NCCL weight broadcast layer parsing
#1996 opened Mar 9, 2026 by zch42 Loading…
1
chore: consolidate vLLM args into [vllm] config section
#1994 opened Mar 9, 2026 by mikasenghaas Loading…
5 tasks done
10
fp8 broadcast
#1981 opened Mar 7, 2026 by S1ro1 Loading…
14
chore: extract configs into standalone prime-rl-configs sub-package
#1977 opened Mar 7, 2026 by samsja Loading…
5 tasks done
2
fp32 chunk
#1971 opened Mar 6, 2026 by S1ro1 Loading…
3
[wip] disagg
#1967 opened Mar 6, 2026 by S1ro1 Loading…
4
top-k masking
#1965 opened Mar 6, 2026 by hallerite Loading…
3
feat: FP8 block-wise online quantization for vLLM inference
#1962 opened Mar 6, 2026 by samsja Loading…
6 of 8 tasks
5
ProTip! Add no:assignee to see everything that’s not assigned.