So in my latest model which is based off movie conversions it was looking promising because in the tests every 1000 steps were looking like it was learning well. But when it finished and I ran interface if was a complete mess, didn't even return the save answers when I copied the examples from the in training, testing every 1000 steps.
Any ideas?
I can provide the train and test files since there are realitivly small if u want.
So in my latest model which is based off movie conversions it was looking promising because in the tests every 1000 steps were looking like it was learning well. But when it finished and I ran interface if was a complete mess, didn't even return the save answers when I copied the examples from the in training, testing every 1000 steps.
Any ideas?
I can provide the train and test files since there are realitivly small if u want.