UForm-v2: tiny CLIP-like embeddings in 21 languages and Graphcore API

Vision-Language understanding Transformer, which has 40% fewer parameters than vanilla CLIP while performing much better on text-to-image retrieval, where it's also beneficial that output embeddings have 2x fewer dimensions (256 vs 512).

Moreover, it supports 21 languages, including popular English, Hindi, Chinese, Arabic, and lower-resource languages like Ukrainian, Hebrew, and Armenian.

Demo: [http://usearch-images.com/](http://usearch-images.com/)

Github: [https://github.com/unum-cloud/uform](https://github.com/unum-cloud/uform)

https://i.redd.it/3zhc95uvs1jb1.gif

api chinese clip datascience embeddings english graphcore hindi image language languages language understanding popular retrieval text text-to-image transformer understanding vision