Stochastic Vehicle Routing

Visually debugging the solution of the model

This branch provides a visual debugger (visual_debugger.ipynb) to compare the solution of the model with the ground truth.

How to use the visual debugger

Run the cells in the notebook following the steps below:

1. Set the arguments in the config file. The default dataset path is data/test.pkl and the default config file is configs/test.yaml.
2. Create the dataset using dataset_creator.py, for example python3 src/dataset_creator.py --city --n_samples 5 --n_tasks 4 --out_file data/test.pkl. The --city flag indicates that the city instance related to a datapoint is stored in the dataset as well. If you use this flag, you have to set city: true in the data args. Without the --city flag, you will not see the visualized solution.
3. Run the trainer code in the notebook, and open the tensorboard using tensorboard --logdir=runs to view the solution.