印度尼西亚BERT2BERT摘要模型

通过使用BERT-base和GPT2-small对印度尼西亚文本进行微调的EncoderDecoder模型用于摘要。

微调语料库

bert2gpt-indonesian-summarization模型基于cahya/bert-base-indonesian-1.5G和cahya/gpt2-small-indonesian-522M，由 cahya 进行微调，使用 id_liputan6 数据集。

加载微调模型

from transformers import BertTokenizer, EncoderDecoderModel

tokenizer = BertTokenizer.from_pretrained("cahya/bert2gpt-indonesian-summarization")
tokenizer.bos_token = tokenizer.cls_token
tokenizer.eos_token = tokenizer.sep_token
model = EncoderDecoderModel.from_pretrained("cahya/bert2gpt-indonesian-summarization")

代码示例

from transformers import BertTokenizer, EncoderDecoderModel

tokenizer = BertTokenizer.from_pretrained("cahya/bert2gpt-indonesian-summarization")
tokenizer.bos_token = tokenizer.cls_token
tokenizer.eos_token = tokenizer.sep_token
model = EncoderDecoderModel.from_pretrained("cahya/bert2gpt-indonesian-summarization")

# 
ARTICLE_TO_SUMMARIZE = ""

# generate summary
input_ids = tokenizer.encode(ARTICLE_TO_SUMMARIZE, return_tensors='pt')
summary_ids = model.generate(input_ids,
            min_length=20,
            max_length=80, 
            num_beams=10,
            repetition_penalty=2.5, 
            length_penalty=1.0, 
            early_stopping=True,
            no_repeat_ngram_size=2,
            use_cache=True,
            do_sample = True,
            temperature = 0.8,
            top_k = 50,
            top_p = 0.95)

summary_text = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
print(summary_text)

输出：