Welcome to OFA-Sys! I hope you enjoy our multimodal models and relevant spaces!

We aim for building a unified multimodal multitask AI system. Toward this goal, we have recently developed a series of multimodal pretrained models, e.g., OFA, Chinese-CLIP, M6, etc. Notably, OFA is a step towards "One For All", as it is a unified multimodal pretrained model that can transfer to a number of downstream tasks effectively (SOTA performance in 2022). For more information, feel free visit our github repo: https://github.com/OFA-Sys

Under this organization, we provide demos of OFA through HF Spaces, and also our models adaptive to HF Transformers.