err when calculate loss in val_op in trainer

qinnzou / DeepCrack

DeepCrack: Learning Hierarchical Convolutional Features for Crack Detection

253 stars 60 forks source link

err when calculate loss in val_op in trainer #18

Closed moxx799 closed 1 year ago

moxx799 commented 2 years ago

hi,

I saw a error in val_op when calculate the loss, you take loss devided by cfg.train_batch_size(which is defined as 8 in cfg), but this step is calculate the validation set, so it must be devided by cfg.val_batch_size.

qinnzou commented 1 year ago

thank you @moxx799