-
Notifications
You must be signed in to change notification settings - Fork 2.9k
Pull requests: huggingface/trl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[GRPO] Apply the completion mask elementwise in the luspo loss aggregation
#6654
opened Aug 3, 2026 by
YaseenBashaT
Loading…
3 of 8 tasks
[DistillationTrainer refactor] Docs: announce the DistillationTrainer promotion in What's New
#6653
opened Aug 3, 2026 by
qgallouedec
Member
•
16/16
Loading…
Allow vLLM weight-update group re-init after a client crash (#3408)
#6651
opened Aug 3, 2026 by
behroozazarkhalili
Collaborator
Loading…
[DistillationTrainer refactor] Support Vision Language Models
#6650
opened Aug 3, 2026 by
qgallouedec
Member
•
14/16
Loading…
[GRPO] Fix entropy bonus normalization inconsistency across loss types
#6648
opened Aug 3, 2026 by
YaseenBashaT
Loading…
3 of 8 tasks
[DistillationTrainer refactor] Move the distillation tests to
tests/
#6647
opened Aug 3, 2026 by
qgallouedec
Member
•
10/16
Loading…
[DistillationTrainer refactor] Re-point ServerDistillationTrainer at the stable base
#6645
opened Aug 3, 2026 by
qgallouedec
Member
•
13/16
Loading…
[DistillationTrainer refactor] Update the example script for the stable API
#6644
opened Aug 3, 2026 by
qgallouedec
Member
•
12/16
Loading…
[DistillationTrainer refactor] Docs: de-namespace DistillationTrainer for the stable API
#6643
opened Aug 3, 2026 by
qgallouedec
Member
•
11/16
Loading…
[DistillationTrainer refactor] Add the
trl distillation CLI
#6642
opened Aug 3, 2026 by
qgallouedec
Member
•
9/16
Loading…
[DistillationTrainer refactor] Move DistillationTrainer to the Stable telemetry group
#6641
opened Aug 3, 2026 by
qgallouedec
Member
•
8/16
Loading…
[DistillationTrainer refactor] Promote DistillationTrainer to the stable API
#6640
opened Aug 3, 2026 by
qgallouedec
Member
•
7/16
Loading…
[DistillationTrainer refactor] Docs: align distillation_trainer.md with stable-trainer conventions
#6639
opened Aug 3, 2026 by
qgallouedec
Member
•
6/16
Loading…
Fix conversational format detection for empty message lists
#6638
opened Aug 3, 2026 by
cupkk
Loading…
4 of 8 tasks
fix(grpo): clip the KL log-ratio before exp to avoid inf overflow
#6637
opened Aug 3, 2026 by
behroozazarkhalili
Collaborator
Loading…
fix(minillm): compute the discounted advantage with a stable reverse recurrence
#6635
opened Aug 3, 2026 by
behroozazarkhalili
Collaborator
Loading…
[DistillationTrainer refactor] Rewrite the test suite to GRPO shape
#6634
opened Aug 3, 2026 by
qgallouedec
Member
•
5/16
Loading…
[DistillationTrainer refactor] Log the student entropy metric
#6633
opened Aug 3, 2026 by
qgallouedec
Member
•
4/16
Loading…
[DistillationTrainer refactor] Config: GRPO parity drift fixes
#6632
opened Aug 2, 2026 by
qgallouedec
Member
•
3/16
Loading…
[DistillationTrainer refactor] Convention/cosmetic sweep
#6630
opened Aug 2, 2026 by
qgallouedec
Member
•
2/16
Loading…
[DistillationTrainer refactor] Reject
use_liger_kernel with logit_scale / final_logit_softcapping
#6629
opened Aug 2, 2026 by
qgallouedec
Member
•
1/16
Loading…
Fix vLLM server hangs after worker errors
#6628
opened Aug 2, 2026 by
DaoyuanLi2816
Contributor
Loading…
5 of 8 tasks
Fix missing entropy gradient in _ChunkedLogProbFunction backward
#6625
opened Aug 2, 2026 by
verma8076
Contributor
Loading…
4 tasks done
Fix FSDP2 vLLM weight sync with ignored parameters
#6623
opened Aug 2, 2026 by
Rusheel86
Loading…
3 of 8 tasks
Previous Next
ProTip!
no:milestone will show everything without a milestone.