token masking and contrastive learning

[ooochen-30]

I wonder how to employ token masking on encoder without decoder in pipeline. Besides, is contrative learning and token masking are used at the same time? can u tell me the detailed operation, thank you.