Could you provide some data samples and arguments to run the code?

[chanhee0222]

Hi,
First of all, awesome work! I really love the paper.

I was trying to reproduce the results from the paper using this repo, but having some trouble.

It seems you have preprocessed the PersonaChat data as well as the feedback data (from Hancock et al.) into your own format, but I can't seem to find any preprocessing code. Could you share an example of the preprocessed data, so I could try to preprocess the data the same way? Releasing the preprocessing script would be also great.

Also, I have noticed that some hyperparameters such as batch size and number of training epochs are missing, both in the paper and in this repo. I guess it needs to be provided as an argument to the main.py script. Could you give me an example of running that script?

I hope I am not troubling you too much.
Thank you in advance.

[ekunnii]

Hello,

Thanks for the feedback.
The codebase is incomplete for now. I have the plan to release the complete codebase in May.

I will address your request this week and push the code.

1. preprocessing notebook
2. preprocessed data
3. training notebook with hyperparameter

[chanhee0222]

That's great! Thank you. Looking forward to it.