Open HHuiwen opened 3 years ago
Hi authors, Thank you very much for sharing the code! In the paper, consistency loss is used for all available samples, but Why you only compute unlabeled data in the code? Is there any wrong with train_mlmt.py?
Hi authors, Thank you very much for sharing the code! In the paper, consistency loss is used for all available samples, but Why you only compute unlabeled data in the code? Is there any wrong with train_mlmt.py?