nllb-200 本地部署

使用facebook的开源翻译模型

nllb-200-distilled-1.3B

测试一下

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

# 加载预训练的分词器和模型
tokenizer = AutoTokenizer.from_pretrained("facebook/nllb-200-distilled-1.3B", use_auth_token=True, src_lang="ron_Latn")
model = AutoModelForSeq2SeqLM.from_pretrained("facebook/nllb-200-distilled-1.3B", use_auth_token=True)

# 要翻译的罗马尼亚语文本
article = "Şeful ONU spune că nu există o soluţie militară în Siria"

# 使用分词器对文本进行编码，将文本转换为模型输入所需的张量格式
inputs = tokenizer(article, return_tensors="pt")

# 生成翻译的令牌序列，强制生成英语 (eng_Latn) 作为目标语言
translated_tokens = model.generate(
     **inputs, forced_bos_token_id=tokenizer.lang_code_to_id["eng_Latn"], max_length=30
 )

# 将生成的令牌序列解码为可读的文本
translated_text = tokenizer.batch_decode(translated_tokens, skip_special_tokens=True)[0]

# 打印翻译结果
print(translated_text)

开源翻译大模型_开源翻译模型-CSDN博客