Resolved: Feature/#80

[seoulsky-field]

Motivation 🤔

• Checking num_classes=5 or num_label=5.
• Both train and evaluation use AUROC metric, however, they use different metric logger.
  

Key Changes 🔑

• Replace num_classes=CONSTANT to num_classes=hydra_cfg.num_classes
• In trainer, we replace existed method to AUROCMetricReporter. (used in inference.py)
• In custom_metrics.py, remove not used methods and existed trainer metric reporter.
• Current working image:

To Reviewers 🙏

• Although this metric logger also uses in conditional_train.py, it might not work well because of #81 .

resolves: #79, #80 references: #51

[seoulsky-field]

@chrstnkgn Thanks a lot! Maybe the conflicts would not be so much, you just consider c_val() function! (add hydra_cfg parameter, change AUROC metric reporter)

@kdg1993 Ooh, it was my mistake. However, in my opinion, it's not necessary. I think our target to use is 1) programmer, 2) doctors, 3) companies. (sorted by importance) Augmentation would be touched by programmer mostly, and they already know how to use Albumentations. So, I thought it's not necessary comment.

Also, it's useful for newbies or people who have not experienced Albumentations, I return that comments.

[kdg1993]

Thanks a lot, @seoulsky-field! I agree with your opinion and thanks for returning it