Thanks to visit codestin.com
Credit goes to github.com

Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: PaddlePaddle/PaddleNLP
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: develop
Choose a base ref
...
head repository: PaddlePaddle/PaddleNLP
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: dsv3-sft
Choose a head ref
  • 14 commits
  • 20 files changed
  • 6 contributors

Commits on Aug 19, 2025

  1. Dsv3 sft (#10968) 

    * update expert parallel init logic

* fix flash_mask && MoEFlexTokenLayer experts && add some config

* offload optimizer

---------

Co-authored-by: blacksheep-Aristotle <[email protected]>
    @deepllz @blacksheep-Aristotle
    deepllz and blacksheep-Aristotle authored Aug 19, 2025
    Configuration menu
    Copy the full SHA
    3794aa9 View commit details
    Browse the repository at this point in the history

Commits on Aug 20, 2025

  1. fix use_rms_norm && add subbatch_token_num config (#10974) 

    Co-authored-by: Your Name <[email protected]>
    @deepllz
    deepllz and Your Name authored Aug 20, 2025
    Configuration menu
    Copy the full SHA
    f2e43b9 View commit details
    Browse the repository at this point in the history

Commits on Aug 22, 2025

  1. moelayer with subbatch to reduce memory (#10985) 

    Co-authored-by: deepllz <[email protected]>
    @deepllz
    deepllz and deepllz authored Aug 22, 2025
    Configuration menu
    Copy the full SHA
    7d5eb9a View commit details
    Browse the repository at this point in the history

Commits on Aug 28, 2025

  1. support sequence parallel in deepseek v3 model 

    * support sequence parallel in deepseek v3

* polish, remove 'print' command
    @pkuzyc
    pkuzyc authored Aug 28, 2025
    Configuration menu
    Copy the full SHA
    4493f19 View commit details
    Browse the repository at this point in the history

  2. enable master_grad && stage1 V2 opt and polish some code

    @deepllz
    deepllz authored Aug 28, 2025
    Configuration menu
    Copy the full SHA
    546e1cb View commit details
    Browse the repository at this point in the history

Commits on Aug 29, 2025

  1. fix some bugs

    @Difers
    Difers authored Aug 29, 2025
    Configuration menu
    Copy the full SHA
    adc2f36 View commit details
    Browse the repository at this point in the history

  2. add warm load (#11029)

    @Difers
    Difers authored Aug 29, 2025
    Configuration menu
    Copy the full SHA
    d1a3d88 View commit details
    Browse the repository at this point in the history

  3. fix bug in subbatch and add 128k sft config (#11032)

    @pkuzyc
    pkuzyc authored Aug 29, 2025
    Configuration menu
    Copy the full SHA
    991a573 View commit details
    Browse the repository at this point in the history

Commits on Sep 1, 2025

  1. compatible with lastest paddle develop branch && update SFT train con… 

    …fig to get better performance
    @deepllz
    deepllz committed Sep 1, 2025
    Configuration menu
    Copy the full SHA
    7adac11 View commit details
    Browse the repository at this point in the history

  2. update dsv3 SFT 128K train config to get better performance (#11037)

    @deepllz
    deepllz authored Sep 1, 2025
    Configuration menu
    Copy the full SHA
    a0fcbdc View commit details
    Browse the repository at this point in the history

Commits on Sep 3, 2025

  1. adapt global_norm_clip for hybrid expert parallel (#11061)

    @pkuzyc
    pkuzyc authored Sep 3, 2025
    Configuration menu
    Copy the full SHA
    ad9e95b View commit details
    Browse the repository at this point in the history

  2. fix aux_loss_alpha && lr value too big problem and add aux update cal… 

    …lback and add mtp subatch_recompute (#11062)

* fix ep grad

* fix aux_loss_alpha && lr value too big problem and add aux update callback and add mtp subatch_recompute

* fix logger error
    @deepllz
    deepllz authored Sep 3, 2025
    Configuration menu
    Copy the full SHA
    7e317e6 View commit details
    Browse the repository at this point in the history

Commits on Sep 5, 2025

  1. fix ep grad bug (#11072)

    @deepllz
    deepllz authored Sep 5, 2025
    Configuration menu
    Copy the full SHA
    6e67781 View commit details
    Browse the repository at this point in the history

Commits on Sep 11, 2025

  1. [Bug Fix]reduce grad of kv_a_proj_with_mqa and q_a_proj to maintain c… 

    …orrectness (#11085)
    @deepllz
    deepllz authored Sep 11, 2025
    Configuration menu
    Copy the full SHA
    7adc457 View commit details
    Browse the repository at this point in the history
Loading