Change --to_csv, --csv_dir to be more user friendly

[ehuan2]

Instead of having it run separately afterwards, it should be a single command and output.

[aosakwe]

For this - right now to_csv lets them specify the metric class and the csv_dir provides the path. In most cases they should just be using the default path so this should be okay as is no?

[ehuan2]

Oh I should specify, if people want the csv right now, they have to first run without it, and then run --to_csv, --csv_dir afterwards themselves for it to run properly. Because the first time is to run the model, the second time is to produce the results.

[aosakwe]

yeah this will compliment #117 as well
although it could just be handled/clarified in an error if no .db file is available

[ehuan2]

Oh so I think what Vishvak wanted was to just run it once, i.e. --to_csv should run the model AND then produce the results.
Actually yeah it should be good to compliment #117 but we can probably do that at a later point in time (would require a little bit of refactoring).

[aosakwe]

Ahh I see but that's what we already get if --config and --to_csv are specified

if a model is being run the config flag has to be called regardless

[ehuan2]

Oh no so what happens if --to_csv is specified is that we don't run the model at all right now, see the main.py, I stop execution early.