Add more models for summarization taskflow

[gongel]

PR types

optimization

PR changes

APIs

Description

• Add more models: IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese-V1, PaddlePaddle/Randeng-Pegasus-238M-Summary-Chinese-SSTIA, PaddlePaddle/Randeng-Pegasus-523M-Summary-Chinese-SSTIA,
• try to fix [Question]: 对于摘要的这个预训练大模型更改配置来替换模型时，会报出错误 #4907
• Use like

from paddlenlp import  Taskflow
summarizer = Taskflow("text_summarization", model="IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese-V1")

[codecov]

Codecov Report

Merging #4933 (2d8d78c) into develop (75f3130) will decrease coverage by 0.01%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           develop    #4933      +/-   ##
===========================================
- Coverage    46.35%   46.34%   -0.01%     
===========================================
  Files          448      448              
  Lines        64643    64646       +3     
===========================================
- Hits         29964    29963       -1     
- Misses       34679    34683       +4     

Impacted Files	Coverage Δ
paddlenlp/taskflow/taskflow.py	80.45% <ø> (ø)
paddlenlp/trainer/trainer.py	59.84% <0.00%> (-0.31%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[linjieccc]

@gongel Hi, 当前还少了两个步骤

1. 配置模型相关文件resource_files_names及下载链接resource_files_urls，在__init__中使用self._check_task_files()触发文件下载

可以参考这里的写法：https://github.com/PaddlePaddle/PaddleNLP/blob/develop/paddlenlp/taskflow/feature_extraction.py#L70

2. 改为from_pretrained(self._task_path)来支持加载微调之后的模型
   https://github.com/PaddlePaddle/PaddleNLP/blob/develop/paddlenlp/taskflow/text_summarization.py#L83

self._model = AutoModelForConditionalGeneration.from_pretrained(self._task_path)

[gongel]

经讨论：Q2升级完后再做统一修改。

[linjieccc]

LGTM