Add caption text decoder (encoder + decoder = full OpenVision2 generative model)

Upload standalone vision encoder: vit-only first commit

initial commit