Difference with official code - utils.MentorNet.forwawrd

[Schmerle-Cki]

Hello, thanks for your reproduction! In utils.py, line 144, upper_bound = torch.where(cur_epoch < burn_in_epoch - 1, v_zeros, v_ones).to(self.device), while in google_mentornet, codes accordingly in utils.py, line 211-212, upper_bound = tf.cond(cur_epoch < (burn_in_epoch - 1), lambda: v_ones, lambda: v_zeros). The order of v_ones and v_zeros seems to be a conflict. Is there a typo, or maybe I have misunderstood some of your definitions？

[Furyton]

Hi, I'm afraid that is indeed a typo. I fixed it in another private repo and forgot to update it here :(
I have fixed it just now. Thank you for your reporting. :)