deepseek-ai

/

DeepSeek-OCR

like 2.28k

Follow

DeepSeek 104k

This PR enhances the model card for DeepSeek-OCR by:

• Adding library_name: transformers to the metadata, which enables the automated "how to use" widget on the model page. This is supported by explicit transformers usage in the inference code and config.json.
• Refining the pipeline_tag from image-text-to-text to image-to-text, which more accurately reflects the model's primary OCR functionality as described in the paper abstract and aligns with Hub conventions.
• Adding an explicit link to the Hugging Face paper page DeepSeek-OCR: Contexts Optical Compression in the navigation links and making the model title clickable to this page. The existing PDF link is renamed to "PDF Paper Link" for clarity.
• Adding a dedicated "Project Page" section with the URL https://www.deepseek.com/ for better discoverability.

These changes aim to improve the model's visibility, usability, and documentation on the Hugging Face Hub.