Add relevant info to model card
Browse filesThis PR:
- links the model to the relevant paper, [PP-DocBee: Improving Multimodal Document Understanding Through a Bag of Tricks](https://huggingface.co/papers/2503.04065)
- ensures the model can be found at https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending
- adds a relevant `library_name`, ensuring the "how to use" button appears on the top right of the model page
- adds a link to the Github code
README.md
CHANGED
|
@@ -1,8 +1,15 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
| 4 |
# PP-DocBee2
|
| 5 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
## 1. 简介
|
| 7 |
|
| 8 |
PP-DocBee2 是PaddleMIX团队自研的一款专注于文档理解的多模态大模型，在PP-DocBee的基础上，我们进一步优化了基础模型，并引入了新的数据优化方案，提高了数据质量，使用自研[数据合成策略](https://arxiv.org/abs/2503.04065)生成的少量的47万数据便使得PP-DocBee2在中文文档理解任务上表现更佳。在内部业务中文场景类的指标上，PP-DocBee2相较于PP-DocBee提升了约11.4%，同时也高于目前的同规模热门开源和闭源模型。
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
library_name: transformers
|
| 4 |
+
pipeline_tag: image-text-to-text
|
| 5 |
---
|
| 6 |
+
|
| 7 |
# PP-DocBee2
|
| 8 |
|
| 9 |
+
This repository contains the model presented in the paper [PP-DocBee: Improving Multimodal Document Understanding Through a Bag of Tricks](https://huggingface.co/papers/2503.04065).
|
| 10 |
+
|
| 11 |
+
For the codebase, see: this https URL
|
| 12 |
+
|
| 13 |
## 1. 简介
|
| 14 |
|
| 15 |
PP-DocBee2 是PaddleMIX团队自研的一款专注于文档理解的多模态大模型，在PP-DocBee的基础上，我们进一步优化了基础模型，并引入了新的数据优化方案，提高了数据质量，使用自研[数据合成策略](https://arxiv.org/abs/2503.04065)生成的少量的47万数据便使得PP-DocBee2在中文文档理解任务上表现更佳。在内部业务中文场景类的指标上，PP-DocBee2相较于PP-DocBee提升了约11.4%，同时也高于目前的同规模热门开源和闭源模型。
|