FISH bot is a generic chat bot that is a sequentional model build using a Bidirection Recurring Neural Netowrk. This chatbot is developed from Daniel Kukiela. Changes have been made to few files to make it domain specific aswell.
- Model trained on Linux machine with 8GB RAM, 8GB P4000 Nvedia GPU and 8 core processor.
- Total number of steps trained: 75K
- 2.5M pairs of conservations from reddit May 2018 comments.
- 125,000 vocab size.
- Domain specific bot trained from data in convo folder.