Mistake in pre-processing

[bnaman50]

Hey,

I believe there is some mistake in pre-processing function (here). As you have yourself mentioned in the comments, you should add [SEP] [CLS] but in the code (line 18), you are adding [CLS] [SEP]

I also checked the training code and it seems they are also doing the same.

This is also verified by looking at the tokens of the datasets (given in training repo). Here {101: '[CLS]', 102: 'SEP'}.

Could you please confirm it?

Thanks, Naman