Hugging Face's logo

HIT-TMG
/
GlyphBERT

Fill-Mask
Transformers
Chinese
bert
bert-base-chinese
Inference Endpoints
Model card Files Community

This project page is about the pytorch code implementation of GlyphBERT by the HITsz-TMG research group.

img.png

GlyphBERT is a Chinese pre-training model that includes Chinese character glyph features.It renders the input characters into images and designs them in the form of multi-channel location feature maps, and designs a two-layer residual convolutional neural network module to extract the image features of the characters for training.

The experimental results show that the performance of the pre-training model can be well improved by fusing the features of Chinese glyphs. GlyphBERT is much better than BERT in multiple downstream tasks, and has strong transferability.

For more details about using it, you can check the github repo

Downloads last month
11
Inference Examples
Fill-Mask
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.