One paper [CAE v2] is accepted by TMLR 2023 on Large-sclae Self-superivised Foundation Model Pretraining