Ha0Tang / AttentionGAN

AttentionGAN for Unpaired Image-to-Image Translation & Multi-Domain Image-to-Image Translation
Other
646 stars 97 forks source link

Checkpoints Missing After Successful Training #23

Closed ghost closed 3 years ago

ghost commented 3 years ago

I used the repo to work on horse2zebra dataset as a test. However, after finishing 60 epochs, although the logs (see below) said multiple .pt files were saved to ./chackpoints, nothing is actually shown there.

saving the model at the end of epoch 60, iters 80160
End of epoch 60 / 60     Time Taken: 444 sec

I used exactly the same command line in scripts/train_attentiongan.sh -- basically copied it to Colab.

I briefly checked the code and believe the checkpoints should have been saved to the default path, because I didn't specify the flags, just like in scripts/train_attentiongan.sh.

Did anyone encountered the same problem?

Ha0Tang commented 3 years ago

please refer to this line

ghost commented 3 years ago

It was probably a problem of syncing in Colab. I was able to find the checkpoints in its linux command line, but not in the left tree-veiw.