[Image Captioning] Training gets stuck after first iteration

[umerebryx]

Hi,

I have tested your train.py script in captioning tutorial to run over my custom dataset. After some debugging it seems the script is not looping back after the first iteration. This is last line executed by the script and then I don't see any output for hours. Below is the output of the script:

python train.py 
Epoch [0/5], Step [0/22], Loss: 4.7755, Perplexity: 118.5671

Have you tested this on custom dataset? Can I get some pointers what might be wrong here.

Thanks

[yunjey]

Did you use a GPU? Image captioning has a long training time. For Pascal Titan X, it took about 1 hour per epoch. Here you can reduce the log_step.

[umerebryx]

Thanks for the reply. Yes, I am using gpu I will reduce the log steps and update here.