Closed jylee-bcm closed 1 month ago
You can see the diff better here:
Running speed might be not a big problem, but I wonder that the memory usage has a drastically change (in a good way)
You can see the diff better here:
Sorry I don't think this diff is much better
Running speed might be not a big problem, but I wonder that the memory usage has a drastically change (in a good way)
Sorry I already cleaned the trace. But it is not expected to improve any performance, because this change was for readability, not the performance.
Running speed might be not a big problem, but I wonder that the memory usage has a drastically change (in a good way)
Sorry I already cleaned the trace.
But it is not expected to improve any performance, because this change was for readability, not the performance.
That was what you first mentioned - even a simple but different loop structure it could optimize memory usage. You stated this multiple times, not one so I believe checking the memory should happen here.
Previously, the code of feature engineering part was written in a single function with lots of dependency. This PR split the function into several pieces so that you can see which the feature dependency easier.
Output Test
I validated with the demo data that the output is identical.
Speed Test
I concerned the speed could be degenerated due to the increased number of steps But it was not substantial. (2m 15s -> 2m 25s)
Evident Dependency
We can see simpler dependency than before.