[FIX] fix the bug that some gemm config did not be handled

[zeroRains]

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

Bug fixes

PR changes

Models

Description

跑模型时出现GEMM 的invalid configure问题

CUDA_VISIBLE_DEVICES=0,1
python -m paddle.distributed.launch --gpus 0,1 predictor.py --model_name_or_path mistralai/Mixtral-8x7B-Instruct-v0.1 --dtype bfloat16 --quant_type weight_only_int8 --mode dynamic --inference_model --append_attn

原因：

模型执行时需要先进行gemm configure搜索，在获取的candidate_configure里，paddle中的三个gemm configure没有在dispatch_moe_gemm_to_cutlass中进行处理，导致触发PADDLE_FATAL直接中断执行进程。

修复方案：

1. 添加缺失的configure处理
2. 在搜索gemm_configure的相关代码中，用PADDLE_THROW替换PADDLE_FATAL

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 46.91%. Comparing base (6f80ea1) to head (f40de2d).
⚠️ Report is 95 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop   #10627      +/-   ##
===========================================
- Coverage    46.91%   46.91%   -0.01%     
===========================================
  Files          799      799              
  Lines       132454   132460       +6     
===========================================
  Hits         62138    62138              
- Misses       70316    70322       +6     

☔ 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.