This assignment is due before 03:00PM on Friday, February 21, 2020.
Implementing HRED in ParlAi : Assignment 3
The point of the third assignment is to reimplement HRED in ParlAi. The focus of this project is to start to be able to build a model in the ParlAi framework and start to try different datasets.
You should be excited about this assignment! You will be able to chat with your own bot!
Be creative and have fun.
- Implement the HRED model within ParlAi. This involved creating the context vector and manipulating the data loading process. This does not have to be an exact reproduction of HRED. You may want to look at Harshal’s implementation or Julian Serban’s old fork of ParlAi.
- Train the HRED model use any dataset that you’d like.
- Integrate your chatbot with Alexa skill
- Evaluate your chatbot. Here, you should show some qualitative examples as well as train and test loss. Furthermore, try to be creative and see if you can make a metric for how well your chatbot is doing.
- Identify issues with your chatbot and suggest possible solutions and future work.
- Submit homework on Gradescpoe.
Please remember to include a README. If there are too many files that simply point to your github … sorry about Gradescope.