【FT】Fix Fault Tolerance in PaddleFleet

[Xing-lil]

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

Others

PR changes

Others

Description

修复容错系统
1、部分参数build_skip_comm_buffer，zcc中未适配

• e_score_correction_bias 设置了build_skip_comm_buffer，"color": "skip_comm"，为stop_gradient，不在 sharding 通信的 fused buffer 中，不会被加入 param_mappings。
• 修复：ZCC中识别unshard的参数，也保存在param_mappings，标记"buffer_index": "unfused"，save时直接保存。

2、grouped_gemm_experts 中的EMA问题

• grouped_gemm在保存EMA ckpt的时候shape未恢复
• 修复：保存EMA时，将grouped_gemm参数恢复为原来的维度

3、MLA的AOA配置问题

• 模型开启了MLA，最新合入的AOA配置未适配MLA
• 修复：补充了MLA的AOA配置

4、GPTModelProvider缺少to_json_string

• 以GPTModelProvider的方式配置模型，config缺少to_json_string
• 修复：补充to_json_string

验证 ZCC 下接续精度对齐

[paddle-bot]

Thanks for your contribution!

[codecov-commenter]

Codecov Report

❌ Patch coverage is 4.91803% with 58 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@6d7d8fb). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...addleformers/trainer/utils/zero_cost_checkpoint.py	0.00%	25 Missing ⚠️
paddleformers/trainer/trainer_utils.py	0.00%	14 Missing ⚠️
paddleformers/transformers/gpt_provider.py	21.42%	11 Missing ⚠️
paddleformers/transformers/minimax_m2/modeling.py	0.00%	8 Missing ⚠️

❌ Your patch status has failed because the patch coverage (4.91%) 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    #4264   +/-   ##
==========================================
  Coverage           ?   33.78%           
==========================================
  Files              ?      474           
  Lines              ?    89370           
  Branches           ?        0           
==========================================
  Hits               ?    30191           
  Misses             ?    59179           
  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.

[wtmlon]

lgtm