tasks.md

ChiNet

NLU project 2 - Story Cloze Task

Source paper: Conditional Generative Adversarial Networks for Commonsense Machine Comprehension (Chinese et al, 2017)

Tasks

[Put your name next to tasks you are currently working on and remove tasks once you have pushed to repo]

• Implement attention

• Xander Implement generator

• Train model (DIFFICULT!)

• Implement result writer

• Write report

Deadlines

Friday 25: Implement disciminator

Friday 1: Implement attention and generator

Tuesday 5: Have model trained and results ready

Thursday 7: Finish report

Friday 8: Hand in project

Structure

• datasets/ - all data sources required for training/validation/testing.
• outputs/ - any output for a model will be placed here, including logs, summaries, checkpoints, and Kaggle submission .csv files.
• src/ - all source code.
  • core/ - base classes
  • datasources/ - routines for reading and preprocessing entries for training and testing
  • models/ - neural network definitions
  • util/ - utility methods
  • main.py - training script

Creating your own model

Model definition

To create your own neural network, do the following:

1. Make a copy of src/models/example.py. For the purpose of this documentation, let's call the new file newmodel.py and the class within NewModel.
2. Now edit src/models/__init__.py and insert the new model by making it look like:

from .example import ExampleNet
from .newmodel import NewModel
__all__ = ('ExampleNet', 'NewModel')

3. Lastly, make a copy or edit src/main.py such that it imports and uses class NewModel instead of ExampleNet.

Training the model

If your training script is called main.py, simply cd into the src/ directory and run

python3 main.py

[The skeleton of this project has been done by Seonwook Park and has been adapted by Nil Adell for this project]