[DATASET] Add cnn_dailymail dataset

[gongel]

PR types

New features

PR changes

Others

Description

• Add cnn_dailymail dataset
• Use like this:

from paddlenlp.datasets import load_dataset
train_set = load_dataset("cnn_dailymail",  splits=["train"]) # version defaults to "3.0.0"
train_set, dev_set, test_set = load_dataset("cnn_dailymail",  splits=["train", "dev", "test"], version="3.0.0")

[ZeyuChen]

Is it possible just use the latest 3.0.0 version instead of involve tfds concept?

[ZeyuChen]

Don't envolve tfds_version concept to paddlenlp dataset, is it possible just use the latest version?

[gongel]

I think paddlenlp should be consistent with the original data. For ease of use, version defaults to "3.0.0".

[smallv0221]

这里还是用version吧

[gongel]

这里还是用version吧

Done, thx

[smallv0221]

我测试一下没读到数据，你测试过了么

[gongel]

我测试一下没读到数据，你测试过了么

我这边测试没问题，第一次加载解压比较慢。

[smallv0221]

LGTM