Open asnota opened 3 years ago
Thanks for your feedbacks! Yes, the hypermeters are quite sensitive, and we are working on improving the robustness of the model right now. I would love to help with your current experiment. Some hyperparameters we were using includes: noi_decay = 1, reduce_decay = 1, prevent = True, reduce_stop = True, lessrepeat = True, Do you mind to share which dataset you have used, what's your hyperparameter configurations and what does the generation look like so that I can help troubleshooting for you? Please expect some delay of my next response but I will try to get back to you ASAP.
Thank you for this information, it indeed helped and now I have comparable results on my side as well!
Awesome! Glad to see that this works for you!
Hello, I'm trying to reproduce the POINTER model behaviour using provided in the repository checkpoints and keys, however the generated text I'm receiving, is not as attractive as the text examples generated with the deployed demo at http://52.247.25.3:8900/
I wonder what parameters are passed in args in order to get the same generation behaviour as from the http://52.247.25.3:8900/?