This PR improves the model card by:
- Adding the `pipeline_tag: audio-to-audio` to the metadata, enabling easier discovery on the Hugging Face Hub at https://huggingface.co/models?pipeline_tag=audio-to-audio.
- Including `library_name: transformers` in the metadata, as evidence from `config.json` and the model's architecture suggests compatibility with the Hugging Face `transformers` library, which will enable the automated "how to use" widget.
- Adding an explicit link to the project page (`https://freedomintelligence.github.io/EchoX/`) in the top section of the model card.
- Replacing the generic "Usage" section with a comprehensive "Sample Usage" section, including environment setup, model download, and inference commands directly from the official GitHub repository's README, allowing users to quickly get started.
- Correcting the training data size in the "Model Description" from 10k hours to 6k hours, aligning with the paper's abstract and the GitHub README.
These chan…
