It would be great to have a way to know if modifications added to the software improve the performance in terms of identifications/runtime in a longitudinal manner.
I am right now thinking of using DVC and have it be part of the CI toolchain.
Ideally on every release and PR you would run it on a series of pre-defined data (what I use for profiling RN...) and keeps the results.
It would be great to have a way to know if modifications added to the software improve the performance in terms of identifications/runtime in a longitudinal manner.
I am right now thinking of using DVC and have it be part of the CI toolchain.
Ideally on every release and PR you would run it on a series of pre-defined data (what I use for profiling RN...) and keeps the results.