语言:

• 俄语 任务标签: 文本到文本生成 评估指标:
• F1分数 标签:
• 语法错误修正
• GEC
• 俄语

这是针对俄语语法错误修正任务微调后的Multilingual Bart模型版本。

初始化模型方法:

from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
model = MBartForConditionalGeneration.from_pretrained("MRNH/mbart-russian-grammar-corrector")

使用分词器示例:

tokenizer = MBart50TokenizerFast.from_pretrained("MRNH/mbart-russian-grammar-corrector", src_lang="ru_RU", tgt_lang="ru_RU")
input = tokenizer("I was here yesterday to studying",text_target="I was here yesterday to study", return_tensors='pt')

生成文本方法:

output = model.generate(input["input_ids"],attention_mask=input["attention_mask"],forced_bos_token_id=tokenizer_it.lang_code_to_id["ru_RU"])

模型训练采用基于隐藏状态输出h的损失计算:

h.logits, h.loss = model(input_ids=input["input_ids"],
                         attention_mask=input["attention_mask"],
                         labels=input["labels"])