Page 1 of 1

Caught exception in thread: '_training_0'

Posted: Thu Aug 06, 2020 2:36 pm
by jss

hi so when i try to train i get this error
i use the original trainer, 32 batch size, 1000000 integrations and 1 gpu.
specs:
intel core i5-6400
nvidia geforce gtx 1050ti
8gb ram


Re: Caught exception in thread: '_training_0'

Posted: Thu Aug 06, 2020 7:54 pm
by abigflea

Im not great at looking at these logs so someone may contradict me.

This folder

Code: Select all

 /deepfake/model burger 

I think there may be an issue with the space in the folder name. Try changing it to something w/o a space.

Otherwise, If I had to guess, its something with the alignments file.
Read through the guide again and see if you missed a detail.

BTW what model are you trying to run with that 1050ti? Original ?


Re: Caught exception in thread: '_training_0'

Posted: Fri Aug 07, 2020 10:24 am
by torzdf

Space in folder name should be fine.

I suspect that you have at least 1 image (maybe more) in your B set that is not the correct size. All the images in your training folders should be output from the extract process, and should be identical sizes.