MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training
Paper β’ 2311.17049 β’ Published β’ 8
An unmodified mirror of Apple's MobileCLIP-S2 image and text encoders in Core ML format, from apple/coreml-mobileclip.
mobileclip_s2_image.mlpackage: image encoder (256Γ256 β 512-d embedding)mobileclip_s2_text.mlpackage: text encoder (CLIP tokens β 512-d embedding)All credit to Apple Machine Learning Research (paper, code). Released under the Apple Machine Learning Research Model License; see LICENSE.