Testing Original Model

[taylormr]

If we wanted to add our own model's performance to the pr_plot, would we take the following steps?

1. Train original model as we like (on whatever data)
2. Use dev.tsv to tune hyperparameters
3. Use original model's output from test.tsv (which is in one of the given formats) by using the flag --openiefour=my_model_output_of_testtsv.txt when calling carb.py
4. Use dump from carb.py and call pr_plot to graph original model against others

Please let me know if I am misunderstanding something! Thank you.