🚀 HUPD T5-小型摘要模型

该HUPD T5-小型摘要模型在HUPD数据集上进行了微调。它最初在这篇论文中被提出。

如需了解有关哈佛美国专利商标局专利数据集（Harvard USPTO Patent Dataset）的更多信息，请访问项目网站或项目的GitHub仓库。

🚀 快速开始

✨ 主要特性

此模型基于HUPD数据集微调，能够对相关文本进行有效摘要提取。

📦 安装指南

文档未提及具体安装命令，跳过此章节。

💻 使用示例

基础用法

你可以直接使用管道（pipeline）进行掩码语言建模：

from transformers import pipeline
summarizer = pipeline(task="summarization", model="HUPD/hupd-t5-small")

TEXT = "1. An optical coherent receiver for an optical communication network, said optical coherent receiver being configured to receive a modulated optical signal and to process said modulated optical signal for generating an in-phase component and a quadrature component, said in-phase component and said quadrature component being electrical signals, said optical coherent receiver comprising a power adjuster in turn comprising: a multiplying unit configured to multiply said in-phase component by an in-phase gain thereby providing a power-adjusted in-phase component, and to multiply said quadrature component by a quadrature gain thereby providing a power-adjusted quadrature component; and a digital circuit connected between output and input of said multiplying unit and configured to compute: a common gain indicative of a sum of a power of said power-adjusted in-phase component and a power of said power-adjusted quadrature component, and a differential gain indicative of a difference between said power of said power-adjusted in-phase component and said power of said power-adjusted quadrature component; and said in-phase gain as a product between said common gain and said differential gain, and said quadrature gain as a ratio between said common gain and said differential gain. 2. An optical coherent receiver according to claim 1, wherein it further comprises an analog-to-digital unit connected at the input of said power adjuster, said analog-to-digital unit being configured to ..."

summarizer(TEXT)

以下是输出结果：

[{'summary_text': 'An optical coherent receiver for an optical communication network includes a power adjuster and a digital circuit connected between output and input of the multiplying unit and configured to compute a common gain indicative of a sum of the power of an in-phase component and the power-adjusted quadrature component, and the differential gain as a product between the common gain and the diffractive gain.'}]

高级用法

你也可以加载模型并按如下方式使用：

import torch
from transformers import AutoTokenizer, AutoModelWithLMHead
# cuda/cpu
device = 'cuda' if torch.cuda.is_available() else 'cpu'
tokenizer = AutoTokenizer.from_pretrained("HUPD/hupd-t5-small")
model = AutoModelWithLMHead.from_pretrained("HUPD/hupd-t5-small").to(device)

inputs = tokenizer(TEXT, return_tensors="pt").to(device)

with torch.no_grad():
    outputs = model.generate(inputs.input_ids, max_new_tokens=256)

generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)

📚 详细文档

如需更多信息，请查看原始论文。

• 论文：The Harvard USPTO Patent Dataset: A Large-Scale, Well-Structured, and Multi-Purpose Corpus of Patent Applications
• 作者：Mirac Suzgun, Luke Melas-Kyriazi, Suproteem K. Sarkar, Scott Duke Kominers, and Stuart M. Shieber
• BibTeX引用：

@article{suzgun2022hupd,
  title={The Harvard USPTO Patent Dataset: A Large-Scale, Well-Structured, and Multi-Purpose Corpus of Patent Applications},
  author={Suzgun, Mirac and Melas-Kyriazi, Luke and Sarkar, Suproteem K and Kominers, Scott Duke and Shieber, Stuart M},
  journal={arXiv preprint arXiv:2207.04043},
  year={2022}
}

📄 许可证

本模型采用CC BY-SA 4.0许可证。

信息表格