相关疑难解决方法(0)

当我们查看 HuggingFaceHub 模型用法时，langchain有这部分作者不知道如何停止生成，https://github.com/hwchase17/langchain/blob/master/langchain/llms/huggingface_pipeline。 py#L182：

class HuggingFacePipeline(LLM):\n        ...\n    def _call(\n        ...\n        if stop is not None:\n            # This is a bit hacky, but I can\'t figure out a better way to enforce\n            # stop tokens when making calls to huggingface_hub.\n            text = enforce_stop_tokens(text, stop)\n        return text\n

我应该使用什么来将停止标记添加到模板的末尾？

如果我们查看https://github.com/hwchase17/langchain/blob/master/langchain/llms/utils.py，它只是一个正则表达式分割，根据停用词列表分割输入字符串，然后取第一个分区re.split

re.split("|".join(stop), text)[0]\n

让我们尝试从 Huggingface 模型中获取生成输出，例如

from transformers import pipeline\nfrom transformers import GPT2LMHeadModel, AutoTokenizer\n\ntokenizer = AutoTokenizer.from_pretrained(\'gpt2\')\nmodel = GPT2LMHeadModel.from_pretrained(\'gpt2\')\n\ngenerator = pipeline(\'text-generation\', …