You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Update the reduce_dict func in qtcls/utils/misc.py to fix the inconsistency in the order of loss term outputs between distributed training and single-gpu training.
Make mixup an optional operation through a new argument --mixup and correct the drop_last logic accordingly.
Cancel the output of Averaged stats in engine.py after each epoch as it is redundant.
Cancel displaying hints when not using pre-training (a feature available in v0.10.x & v0.11.x). Only provide hints when using pre-training or resuming training.
Provide hints when setting both --pretrain and --no_pretrain.
Provide more hints when the component name (dataset, model lib, model, criterion, optimizer, scheduler, and evaluator) is not found.
Update docs, highlighting the use of config files.