pierre-zhao/RoR

Can not find answer calibration part in code.

Closed this issue · 2 comments

Thanks for your great work.
I'm trying to reproduce your result. All things go well, except the answer calibration part. I could not find related code.
So here are my questions:

  1. where can I find code about answer calibration ?
  2. you put the cls token at the end of the input, any diff from the original star pos?
  3. when pred the end pos, it combines the start hidden state, does it improve the final result?
  4. could you provide some details about transfer learing?

All problems are solved by the author, close this issue.