Dont Know Response

🚀 自然不知道回复模型

本模型基于 Google的T5 进行微调，结合了基于依赖规则的数据和 Quora问题对(QQP) 数据集，用于不知道回复生成任务。

🚀 快速开始

模型信息

相关资料

• 论文：说“不”是一门艺术：不可回答对话查询的上下文回退回复
• Github仓库：https://github.com/kaustubhdhole/natural-dont-know

💻 使用示例

基础用法

from transformers import T5ForConditionalGeneration, T5Tokenizer
model_name = "ashish-shrivastava/dont-know-response"
model = T5ForConditionalGeneration.from_pretrained(model_name)
tokenizer = T5Tokenizer.from_pretrained(model_name)

input = "Where can I find good Italian food ?"
input_ids = tokenizer.encode(input, return_tensors="pt")
outputs = model.generate(input_ids)
decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(decoded_output) # I'm not sure where you can get good quality Italian food.

🔧 技术细节

超参数

n_epochs = 2
base_LM_model = "T5-base"
max_seq_len = 256
learning_rate = 3e-4
adam_epsilon = 1e-8
train_batch_size = 6

📄 许可证

BibTeX引用信息

@misc{shrivastava2020saying,
      title={Saying No is An Art: Contextualized Fallback Responses for Unanswerable Dialogue Queries}, 
      author={Ashish Shrivastava and Kaustubh Dhole and Abhinav Bhatt and Sharvani Raghunath},
      year={2020},
      eprint={2012.01873},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}