TecoGAN icon indicating copy to clipboard operation
TecoGAN copied to clipboard

Published 20 hours ago • verified publisher icon thunil

Training a new TecoGAN model gives an increasing loss

Open Liang-Rui opened this issue 5 years ago • 4 comments

Hi,

Anyone has tried to train a new TecoGAN model? My training processing gives an increasing loss, can you give me any hint by inspecting the log file?

Screen Shot 2019-12-12 at 9 57 40 am Screen Shot 2019-12-12 at 9 58 05 am Screen Shot 2019-12-12 at 9 58 30 am Screen Shot 2019-12-12 at 9 58 48 am Screen Shot 2019-12-12 at 9 59 24 am Screen Shot 2019-12-12 at 9 59 57 am Screen Shot 2019-12-12 at 10 00 13 am Screen Shot 2019-12-12 at 10 00 32 am

logfile.txt

Liang-Rui avatar Dec 12 '19 05:12 Liang-Rui

Excuse me, can you train this model? I have faced a problem that vgg model can't be restored, have you encountered this problem？

Feihong-cc avatar Jan 18 '20 05:01 Feihong-cc

hey @Liang-Rui , how did you print those losses plots?

alessiapacca avatar Oct 26 '20 13:10 alessiapacca

hey @Liang-Rui , how did you print those losses plots? 1.a=your checkpoint /log folder address

  1. replace a with step1 tensorboard --logdir=a --host=127.0.0.1 3, use http://127.0.0.1:6006/ or something that cmd will appear use that address
  2. will see the plots ,enjoy

FrankLinxzx avatar Feb 28 '21 01:02 FrankLinxzx

@Liang-Rui Did you figure out why is there a increase in loss, cause I'm facing the same problem.

santosh-shriyan avatar May 28 '21 07:05 santosh-shriyan