Clover Image Tiny โ€” Core ML

Versioned Core ML resources for the Clover iPhone app.

manifest.json is the public download contract. The iOS 18 release stores Clover's base weights once in a stateful U-Net and can mix up to three rank-16 LoRAs at runtime. Each optional style remains its original 6,927,128-byte safetensors file; it is not a duplicated 600 MB U-Net.

The v4 state layout uses block-concatenated LoRA ranks. This implements the exact weighted sum of the selected styles without cross terms, while unused slots remain zero and produce the ordinary Clover output.

The base checkpoint and style adapters are derivatives of neonforestmist/Clover-Image-Tiny and retain the CreativeML Open RAIL-M license. Style training datasets are Apache-2.0 and linked by the manifest.

Citation

@software{lozadaperez2026cloverimagetinycoreml,
  author = {Lukas Lozada Perez},
  title = {Clover Image Tiny Core ML: On-Device Text-to-Image Resources for iPhone},
  year = {2026},
  url = {https://huggingface.co/neonforestmist/Clover-Image-Tiny-CoreML}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Collection including neonforestmist/Clover-Image-Tiny-CoreML