Trace Captioning datasets collection. The datasets were originally introduced in the paper "One Patch to Caption Them All".