The model can generate definition in Chinese. Usage:

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline

checkpoint = 'PoeticPaper/mbart-large-50_definition_zh_CN'
tokenizer = AutoTokenizer.from_pretrained(checkpoint, src_lang="zh_CN", tgt_lang="zh_CN")
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
generator = pipeline(
    task='text2text-generation',
    model=model,
    tokenizer=tokenizer,
    forced_bos_token_id=tokenizer.lang_code_to_id["zh_CN"],
)

input = "define: 槌 context: 能够 带来 幸福 的 小 槌 吊饰 共有 金白 两色 。 definition:"
output = generator(input) # output will be: "圆柱形 的 物体 。""