8000 [LLM] [Cherry-Pick] Integrate PDC SDK for LLM training fault tolerance platform by SylarTiaNII · Pull Request #9706 · PaddlePaddle/PaddleNLP · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

[LLM] [Cherry-Pick] Integrate PDC SDK for LLM training fault tolerance platform #9706

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 8 commits into from
Jan 3, 2025
Merged

[LLM] [Cherry-Pick] Integrate PDC SDK for LLM training fault tolerance platform #9706

merged 8 commits into from
Jan 3, 2025

Conversation

SylarTiaNII
Copy link
Contributor
@SylarTiaNII SylarTiaNII commented Dec 26, 2024 •
edited
Loading

PR types

New features

PR changes

Others

Description

CP from: #9263; #9431; #9443; #9539; #9618; #9690;

  • support automatically obtaining resume step from FT platform
  • support informing training args to FT platform with json file
  • support paddle recall error for recall & recovery by FT platform
  • support downloading & uploading interactions with remote storage through FT platform agent
  • support deploying freeze model on FLASH device through FT platform agent

[Pcard-88789]

@paddle-bot Paddle-bot
Copy link
paddle-bot bot commented Dec 26, 2024

Thanks for your contribution!

@codecov Codecov
Copy link
codecov bot commented Dec 26, 2024 •
edited
Loading

Codecov Report

Attention: Patch coverage is 17.64706% with 364 lines in your changes missing coverage. Please review.

Project coverage is 52.64%. Comparing base (fa3fd39) to head (9612519).
Report is 269 commits behind head on develop.

Files with missing lines Patch % Lines
paddlenlp/utils/pdc_sdk.py 15.56% 282 Missing ⚠️
paddlenlp/trainer/argparser.py 8.10% 34 Missing ⚠️
paddlenlp/trainer/trainer_utils.py 15.38% 22 Missing ⚠️
paddlenlp/utils/downloader.py 17.64% 14 Missing ⚠️
paddlenlp/trainer/trainer.py 25.00% 6 Missing ⚠️
paddlenlp/utils/fault_tolerance.py 66.66% 4 Missing ⚠️
paddlenlp/trainer/training_args.py 75.00% 2 Missing ⚠️
Additional details and impacted files 
@@             Coverage Diff             @@
##           develop    #9706      +/-   ##
===========================================
+ Coverage    51.11%   52.64%   +1.53%     
===========================================
  Files          730      720      -10     
  Lines       122587   112928    -9659     
===========================================
- Hits         62657    59453    -3204     
+ Misses       59930    53475    -6455

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@luoyuedong @SylarTiaNII
args支持自动保存本地json (PaddlePaddle#9263)
472208f 
* args支持自动保存本地json

* args支持自动保存本地json
@luoyuedong @SylarTiaNII
增加容错的resume-ckpt解析 (PaddlePaddle#9539)
0f6d3f2 
* 增加resume-ckpt支持

* 增加resume-ckpt支持

* 增加resume-ckpt支持

* pdc-versioncontrol-6234 pdc-versioncontrol-6400 sft 多阶段接口

* 增加resume-ckpt支持

* 增加resume-ckpt支持
wawltor
wawltor reviewed Jan 2, 2025
View reviewed changes
paddlenlp/utils/pdc_sdk.py

from paddlenlp.utils.log import logger

PDC_AGENT_BIN = "/root/paddlejob/tools/agent"
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里的PATH设置成固定会不会有坑，如有其他平台或者PDC的其他镜像想使用的时候，不一定有这个path

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个不会的，/root/paddlejob/tools 不是镜像的路径，是PDC所有插件预置的路径。每次拉起容器的时候都会在这个目录里实时部署最新插件组，因此是不会随便改的。

ZHUI
ZHUI approved these changes Jan 2, 2025
View reviewed changes
Copy link
Collaborator
@ZHUI ZHUI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@ZHUI ZHUI merged commit 7713ca8 into PaddlePaddle:develop Jan 3, 2025
8 of 12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@wawltor wawltor wawltor left review comments

@ZHUI ZHUI ZHUI approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@SylarTiaNII @wawltor @ZHUI @D0m021ng @luoyuedong
0