[LLM] Add expert parallel

[DrownFish19]

PR types

New features

PR changes

Models

Description

Add DeepSeekV2.

[paddle-bot[bot]]

Thanks for your contribution!

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 15.04221% with 1107 lines in your changes missing coverage. Please review.

Project coverage is 52.56%. Comparing base (bc55104) to head (decf92b). Report is 41 commits behind head on develop.

:exclamation: Current head decf92b differs from pull request most recent head d49640d

Please upload reports for the commit d49640d to get more accurate results.

Files with missing lines	Patch %	Lines
paddlenlp/transformers/deepseek_v2/modeling.py	14.43%	759 Missing :warning:
paddlenlp/transformers/moe_gate.py	13.33%	221 Missing :warning:
paddlenlp/transformers/moe_layer.py	21.05%	75 Missing :warning:
...addlenlp/transformers/deepseek_v2/configuration.py	13.04%	40 Missing :warning:
...ddlenlp/transformers/deepseek_v2/tokenizer_fast.py	29.41%	12 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## develop #9061 +/- ## =========================================== - Coverage 53.07% 52.56% -0.51% =========================================== Files 656 662 +6 Lines 106095 107614 +1519 =========================================== + Hits 56309 56567 +258 - Misses 49786 51047 +1261 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[CLAassistant]

All committers have signed the CLA.