This was the start of a few ideas I had: splitting benchmarks into separate files, allowing individual command-line runs, etc. and avoiding the need to install vbench, numpy, pandas, sqlalchemy, pytz before being able to run the suite.
If you like this direction, I was also planning to refactor the test data a bit and probably separate repo with some larger test images so the main git repo doesn't get too large.
This was the start of a few ideas I had: splitting benchmarks into separate files, allowing individual command-line runs, etc. and avoiding the need to install vbench, numpy, pandas, sqlalchemy, pytz before being able to run the suite.
If you like this direction, I was also planning to refactor the test data a bit and probably separate repo with some larger test images so the main git repo doesn't get too large.