Nlp 飞马座抽象概要参数

我正在玩pegasus文档中的用法示例。它在collab中运行良好，但使用了提取摘要。有没有办法把这个例子转换成抽象的？示例代码为：

from transformers import PegasusForConditionalGeneration, PegasusTokenizer
import torch
src_text = [
    """ PG&E stated it scheduled the blackouts in response to forecasts for high winds amid dry conditions. The aim is to reduce the risk of wildfires. Nearly 800 thousand customers were scheduled to be affected by the shutoffs which were expected to last through at least midday tomorrow."""
]

model_name = 'google/pegasus-xsum'
torch_device = 'cuda' if torch.cuda.is_available() else 'cpu'
tokenizer = PegasusTokenizer.from_pretrained(model_name)
model = PegasusForConditionalGeneration.from_pretrained(model_name).to(torch_device)
batch = tokenizer.prepare_seq2seq_batch(src_text, truncation=True, padding='longest').to(torch_device)
translated = model.generate(**batch)
tgt_text = tokenizer.batch_decode(translated, skip_special_tokens=True)
assert tgt_text[0] == "California's largest electricity provider has turned off power to hundreds of thousands of customers."

谢谢你的帮助