生成对话标题

如何使用

model_name = "theblackcat102/alpaca-title-generator-mt0-large"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
question = 'Hi\nHow can I help you?'
encodes = tokenizer(question, return_tensors='pt')
outputs = model.generate(encodes.input_ids, 
    max_length=512,
    do_sample=True,
    repetition_penalty=1.2,
    top_k=50,
    num_return_sequences=1,
    early_stopping=True
)
for i, beam_output in enumerate(outputs):
    print('-----')
    print("{}".format(tokenizer.decode(beam_output, skip_special_tokens=True)))
# > Help requested.

生成标题数据

使用 yahma/alpaca-cleaned 的响应对生成数据，并使用openai turbo模型生成标题。

""
user: {}
assistant: {}
""

Generate a very short title within 5 words of the conversation above, title must be as relevant as possible. Title language must be same as the context

TITLE: