Daniel Kukieła

In this case it looks like there's no checkpoint saved yet. Make sure there is a checkpoint saved.

It means that something went wrong. What training data are you using?

It looks like you have used a dictionary file from other dataset, hmmm. Is it possible? Or you were trying to train with that data only?

That's a bit odd situation and i would like to ask couple of questions here: 1. What operating system 2. What version of Python (and 32- or 64-bit? You should...

I'm not sure about that, I was not trying latest version of TF against our chatbot. But i heard about that problem. As for now i can't check.

Are you training it on a CPU? Do you have Nvidia GPU? Demo model should be trained for couple of epochs, but won't be decent on that training data. Repository...

First - sorry for late reply. How much data are you using? Also are you sure you are using TF-GPU?

That means that you have too few data in your training dataset or you have to train it longer. Chatbot is bound somehow to links. You could also remove from...

My test bot is trained on 1 or 2 million pairs iirc for 5 epochs. Give it a try for 1 or 2 more epochs.

Hi, Could you describe your problem more, please? `full_some_questions-81k.out` is nothing my code outputs, and none of data files is related to `scoring.py` as well, so I'm not sure what...