RuiSumida's picture
Upload README.md with huggingface_hub
8f09584 verified
---
license: other
license_name: apple-ascl
license_link: https://github.com/apple/ml-mobileclip/blob/main/LICENSE_weights_data
library_name: coreml
tags:
- mobileclip
- coreml
- clip
---
# MobileCLIP-S0 Core ML (re-hosted for the Luke app)
Apple's official Core ML export of **MobileCLIP-S0** ([apple/coreml-mobileclip](https://huggingface.co/apple/coreml-mobileclip)),
re-packaged as `.mlpackage.tar.gz` archives for direct on-device download + compilation.
- `MobileCLIPImageEncoder.mlpackage.tar.gz` — image encoder (256×256 RGB → 512-dim embedding, output `final_emb_1`)
- `MobileCLIPTextEncoder.mlpackage.tar.gz` — text encoder (1×77 int32 CLIP BPE tokens, zero-padded → 512-dim embedding)
Weights are unchanged from Apple's release and remain under the
[Apple Sample Code License](https://github.com/apple/ml-mobileclip/blob/main/LICENSE_weights_data).
Reference: [MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training](https://arxiv.org/pdf/2311.17049.pdf) (CVPR 2024).