CUDAOUTOFMEMORY

/

CLEAR

like 0

Hi! I'm Niels from the community science team at Hugging Face. I'm opening this PR to add a model card for CLEAR, a unified multimodal model that leverages generative capabilities (image restoration) to improve the visual understanding of degraded images.

This PR includes:

• Metadata with the any-to-any pipeline tag.
• Links to the paper, project page, and GitHub repository.
• A summary of the three-stage training pipeline (SFT, Bridge Training, and Interleaved GRPO).
• Citation information.