-
Notifications
You must be signed in to change notification settings - Fork 499
Pull requests: areal-project/AReaL
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix(ppo): exclude no-eos rows from reward normalization
#1351
opened May 19, 2026 by
haoyang9804
Contributor
Loading…
fix: rollout version dump - filter by loss_mask and add version_rle
#1350
opened May 19, 2026 by
pyq623
Loading…
fix(utils): ignore masked invalid normalization values
safe-to-test
Ready to run unit-tests in a PR.
#1347
opened May 18, 2026 by
haoyang9804
Contributor
Loading…
fix(infra): correct staleness capacity inflation after recovery
#1345
opened May 16, 2026 by
daihaowz
Collaborator
Loading…
5 of 9 tasks
fix(checkpointer): use dp_reshardable sharding type for megatron-core >=0.11
#1344
opened May 15, 2026 by
theNefelibata
Loading…
2 tasks
feat: enable v2 training pipeline with controller parity
#1327
opened May 11, 2026 by
garrett4wade
Collaborator
•
Draft
7 of 11 tasks
feat(examples): add OSWorld GRPO training example
#1326
opened May 11, 2026 by
hehui0226
Loading…
5 of 9 tasks
feat: Support Linear Cross Entropy fuse kernel
#1322
opened May 10, 2026 by
TaoZex
Collaborator
Loading…
5 of 15 tasks
Refined Kubernetes scheduler implementation
#1316
opened May 8, 2026 by
senseipri
Loading…
8 of 15 tasks
ci: add real training jobs to nightly workflow
#1313
opened May 7, 2026 by
garrett4wade
Collaborator
•
Draft
feat(awex): add colocated CUDA IPC weight transfer
#1310
opened May 6, 2026 by
garrett4wade
Collaborator
Loading…
6 of 9 tasks
feat(experimental): integrate Ray RDT for weight syncing
#1305
opened May 6, 2026 by
KaisennHu
Loading…
3 of 9 tasks
feat(archon): add ZERO1 DTA path with configs and tests
#1287
opened Apr 28, 2026 by
ezoicoder
Collaborator
Loading…
7 of 15 tasks
feat: muon optimizer support
#1270
opened Apr 27, 2026 by
HT-Yuan
Collaborator
Loading…
2 of 15 tasks
feat:Support LoRA incremental weight synchronization on disk for FSDP and SGLang
#1233
opened Apr 23, 2026 by
TaoZex
Collaborator
Loading…
2 of 8 tasks
fix: prevent async RL dispatch crashes on uneven batches
reviewed
stale
#1225
opened Apr 22, 2026 by
yyypluto
Loading…
feat(rs): add two-stage Geo-RS + Token-MIS/TIS mode to RejectionSamplingConfig
stale
#1218
opened Apr 20, 2026 by
morgan-heisler
•
Draft
8 tasks done
feat: add router replay(R3) for megatron engine
high priority
#1207
opened Apr 18, 2026 by
TaoZex
Collaborator
Loading…
5 of 15 tasks
feat: support structured reward outputs and grouped reward aggregation
stale
#1200
opened Apr 17, 2026 by
Wangxiaoxiaoa
Contributor
Loading…
3 of 15 tasks
[draft] Elastic weight update setup and acceleration #1101
stale
#1188
opened Apr 15, 2026 by
sjmshsh
Loading…
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.