I'm looking into better testing the performance of our software on large systems, both for generation user experience and to safeguard against regressions. I would like to see the runtime of notebook pre-processing broken down per-notebook (or per-cell).
When pytest is managing notebooks, I can add --durations=20 to get an idea which cells are the slowest. (Usually there's something of a Pareto distribution, <5 cells taking most of the wall time.)
I'm looking into better testing the performance of our software on large systems, both for generation user experience and to safeguard against regressions. I would like to see the runtime of notebook pre-processing broken down per-notebook (or per-cell).
When
pytestis managing notebooks, I can add--durations=20to get an idea which cells are the slowest. (Usually there's something of a Pareto distribution, <5 cells taking most of the wall time.)