jjiaweiyang

/

l-DeTok

like 1

This PR updates the pipeline_tag for the model card from image-feature-extraction to unconditional-image-generation.

This change better reflects the primary application and improved performance of the l-DeTok tokenizer in downstream image generation tasks, as highlighted in the paper's abstract ("consistently outperforms standard tokenizers across six representative generative models"). This ensures the model is discoverable under the most relevant pipeline on the Hub.

The library_name remains pytorch as the official implementation is built in PyTorch.