jhu-clsp

/

ettin-encoder-68m

like 1

Follow

Center for Language and Speech Processing @ JHU 102

This PR improves the model card for the jhu-clsp/ettin-encoder-68m model by updating its metadata and ensuring the content is the most current available.

Key changes include:

• pipeline_tag: Changed from fill-mask to feature-extraction. This more accurately reflects the model's primary use cases, which include classification and retrieval, as demonstrated by the performance benchmarks and usage examples. This will improve its discoverability for users seeking embedding or classification models.
• library_name: Added library_name: transformers. This explicitly states compatibility with the Hugging Face Transformers library, enabling the "Use in Transformers" widget on the model page for improved usability.
• tags: Added modernbert as an architectural tag, as indicated by the model's config.json (ModernBertForMaskedLM) and mentions in the paper and GitHub README. This provides more precise information about the model's lineage.
• Content: The content has been updated to match the more recent and comprehensive GitHub README, which includes an accurate count of "24 Models" and a direct link to the published Arxiv paper.

These updates enhance the model card's accuracy, discoverability, and user experience on the Hugging Face Hub.