-
Notifications
You must be signed in to change notification settings - Fork 3.1k
Pull requests: PaddlePaddle/PaddleNLP
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Translation of Documentation for Release/2.3
contributor
documentation
Improvements or additions to documentation
#2567
opened Jun 19, 2022 by
lilyyjzhou
Loading…
[fix] zcc ema under non-pp when
acc=1
contributor
stale
#9941
opened Feb 25, 2025 by
Meiyim
Loading…
Perform Sharding V1 reshard forcely when param2rank is empty (#9214)
stale
#9260
opened Oct 14, 2024 by
luoyuedong
Loading…
[LLM] add token metrics after regular & save step
stale
#9146
opened Sep 14, 2024 by
SylarTiaNII
Loading…
[LLM] add learning rate customizing interface
stale
#9596
opened Dec 10, 2024 by
SylarTiaNII
Loading…
Setting Param/Grad/OPT sync default =True in TP
stale
#8841
opened Jul 31, 2024 by
ForFishes
Loading…
fix reshard_qkv alltoall error
stale
#9950
opened Feb 27, 2025 by
blacksheep-Aristotle
Loading…
2 tasks
fix rng_state checkpoint error when tp8
stale
#10187
opened Mar 19, 2025 by
blacksheep-Aristotle
Loading…
2 tasks
[LLM INFER] Support dynamic graph server
stale
#10277
opened Mar 27, 2025 by
yuanlehome
•
Draft
2 tasks
moe_layer support fine_grained_forward
#10980
opened Aug 21, 2025 by
blacksheep-Aristotle
Loading…
2 tasks
recompute support offload tensor
#10981
opened Aug 21, 2025 by
blacksheep-Aristotle
Loading…
2 tasks
Add fused_topk_to_multihot / fused_multihot_prob_backto_topk(grad) custom op to prevent CPU stall.
#10125
opened Mar 13, 2025 by
A-nnonymous
Loading…
save args for rms_norm output in backward
#10938
opened Aug 13, 2025 by
chen2016013
Loading…
2 tasks
Previous Next
ProTip!
no:milestone will show everything without a milestone.