loss value

[ciande-narbit]

Follow the script you provided, my training losses fluctuate wildly, unable to convergence. Will your training losses be significantly reduced?

[PeterWang512]

Do you mind sharing which script are you using to train the model? Thanks!

[ciande-narbit]

train_teaser_cat.sh
with loss below

[ciande-narbit]

Here is the content of script

#!/bin/bash
python train.py
--name teaser_cat_augment --batch 4
--dataroot_sketch ./data/sketch/by_author/cat
--dataroot_image ./data/image/cat --l_image 0.7
--g_pretrained ./pretrained/stylegan2-cat/netG.pth
--d_pretrained ./pretrained/stylegan2-cat/netD.pth
--max_iter 150000 --disable_eval --diffaug_policy translation \