Skip to content

Add deepseekv32 model fix#4507

Open
xingmingyyj wants to merge 6 commits into
PaddlePaddle:developfrom
xingmingyyj:add_deepseekv32_model_fix
Open

Add deepseekv32 model fix#4507
xingmingyyj wants to merge 6 commits into
PaddlePaddle:developfrom
xingmingyyj:add_deepseekv32_model_fix

Conversation

@xingmingyyj
Copy link
Copy Markdown
Collaborator

@xingmingyyj xingmingyyj commented May 22, 2026

Before submitting

  • Lint code. If there are lint issues, please format the code first.
# Install and register `pre-commit` in the project folder
pip install pre-commit && pre-commit install

# Process previous code files separately
pre-commit run --file XXXX.py
  • Add test cases into tests folder. If there are codecov issues, please add tests cases first.

PR types

PR changes

Description

@xingmingyyj
add deepseek v3.2 model
783a6d3
@xingmingyyj
add deepseek v3.2 model
63acc2b
@xingmingyyj
fix
3456339
@xingmingyyj
tmp
2c8b687
@xingmingyyj
Merge develop into add_deepseekv32_model_fix
6d8917c 
Conflict resolution:
- aoa_config_base.py: use develop's version entirely
- model_utils.py: keep develop's dtype aoa logic (with is_fleet guard)
- gpt_provider.py: add mscale_all_dim extraction with safe key check
- training_args.py: keep both dsa_indexer_loss_coeff and develop's new fields
- template.py: keep both deepseek_v32 and glm_ocr templates

Adapt deepseek v3.2 to develop's API:
- Rename moe_grouped_gemm -> moe_expert_fusion in provider
- Add multi_latent_attention=True and use_qk_norm=True to config
- Wire up gen_inv_aoa_config from base class
@paddle-bot
Copy link
Copy Markdown

paddle-bot Bot commented May 22, 2026

Thanks for your contribution!

@codecov-commenter
Copy link
Copy Markdown

codecov-commenter commented May 22, 2026

Codecov Report

❌ Patch coverage is 11.57025% with 107 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@a8ab43b). Learn more about missing BASE report.

Files with missing lines Patch % Lines
...addleformers/transformers/deepseek_v32/modeling.py 0.00% 58 Missing ⚠️
...formers/transformers/deepseek_v32/configuration.py 11.32% 47 Missing ⚠️
paddleformers/transformers/gpt_provider.py 33.33% 2 Missing ⚠️

❌ Your patch status has failed because the patch coverage (11.57%) is below the target coverage (75.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files 
@@            Coverage Diff             @@
##             develop    #4507   +/-   ##
==========================================
  Coverage           ?   46.39%           
==========================================
  Files              ?      478           
  Lines              ?    90760           
  Branches           ?        0           
==========================================
  Hits               ?    42108           
  Misses             ?    48652           
  Partials           ?        0

☔ 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.

@Paddle-CI-Bot
Copy link
Copy Markdown

Paddle-CI-Bot commented May 22, 2026 •
edited
Loading

PaddleFormers Log Analysis

Run #26278781942 · Attempt 1

日志分析报告

流水线名称 问题标签 修复建议 日志片段
CI_ILUVATAR 其他（Runner 基础设施错误） 自托管 Runner 容器执行失败，非代码问题，联系 CI 维护人员重跑 报错代码
Codestyle Check 代码风格（black 格式化失败） 本地运行 pre-commit run --file paddleformers/transformers/deepseek_v32/modeling.py paddleformers/transformers/aoa_config_base.py 后提交格式化结果 报错代码

失败的测试case:

CI_ILUVATAR:       ##[error]Error: failed to run script step: [object Object]
                   ##[error]Process completed with exit code 1.
                   ##[error]Executing the custom container implementation failed.

Codestyle Check:   black hook Failed
                   reformatted paddleformers/transformers/deepseek_v32/modeling.py
                   reformatted paddleformers/transformers/aoa_config_base.py
                   make: *** [Makefile:23: lint] Error 1

根本原因分析:
PR #4507（Add deepseekv32 model fix）新增了 paddleformers/transformers/deepseek_v32/modeling.py 和 paddleformers/transformers/aoa_config_base.py，这两个文件未经 black 格式化直接提交，导致 Codestyle Check 失败；CI_ILUVATAR 失败为天数 Runner 自身容器执行异常（failed to run script step: [object Object]），与 PR 代码无关。

修复建议:

  1. Codestyle Check（必须修复）：

    pip install pre-commit && pre-commit install
pre-commit run --file paddleformers/transformers/deepseek_v32/modeling.py paddleformers/transformers/aoa_config_base.py
git add paddleformers/transformers/deepseek_v32/modeling.py paddleformers/transformers/aoa_config_base.py
git commit -m "style: apply black formatting to deepseek_v32 and aoa_config_base"
git push

  2. CI_ILUVATAR（基础设施问题）：Runner 报 Executing the custom container implementation failed，属于天数机器侧问题，直接 Re-run job 即可；若持续失败，联系 CI 维护人员检查 iluvatar-gpu-2-nczzk-runner-lrhqp 节点状态。

🔄 每次 Re-run 后自动更新

@xingmingyyj
Copy link
Copy Markdown
Collaborator Author

xingmingyyj commented May 22, 2026

/re-run all-failed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@xingmingyyj @codecov-commenter @Paddle-CI-Bot