[LLM] support QWen1.5-Moe

[DrownFish19]

PR types

New features

PR changes

Models

Description

1. add QWen1.5 Moe model.
2. support same prefix for different models, such as QWen and QWen2Moe with same prefix QWen. The longest name will match each model name before others.
3. support sft and lora.

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 66.55556% with 301 lines in your changes are missing coverage. Please review.

Project coverage is 54.39%. Comparing base (3aa92ce) to head (58af3ec).
Report is 51 commits behind head on develop.

Files	Patch %	Lines
paddlenlp/transformers/qwen2moe/modeling.py	72.33%	197 Missing ⚠️
paddlenlp/transformers/qwen2moe/tokenizer.py	22.96%	104 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #8338      +/-   ##
===========================================
- Coverage    55.36%   54.39%   -0.98%     
===========================================
  Files          614      621       +7     
  Lines        96016    97254    +1238     
===========================================
- Hits         53164    52904     -260     
- Misses       42852    44350    +1498     

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

[DesmonDay]

QWen2MoEConfig会不会更好，把Moe都改成MoE。

[DesmonDay]

2023 -> 2024，都改掉吧