Repository navigation
large negative increment values in line profiler #195
Description
Activity
Similar issue here. Totals don't add up and the increments are large and negative.
Line # Mem usage Increment Line Contents ================================================ 5 23.891 MiB 23.891 MiB 6 7 8 23.891 MiB 0.000 MiB 9 23.891 MiB 0.000 MiB 10 11 24.137 MiB 0.246 MiB 12 13 14 112.980 MiB 88.844 MiB 15 16 112.980 MiB 0.000 MiB 17 18 136.840 MiB -2654.883 MiB 19 20 136.840 MiB -2676.617 MiBThanks @nathantsoi and @alowayed for the report. This bug doesn't seem to appear in the standard "test" suite (i.e., typing make test from the root directory). Could you paste a minimal reproducing example for this?
Any progress on this?
@karan10111 we would need some help in having a short, self-contained, correct example to reproduce this. Do you have a short snippet that we can use for that?
So, most of the times you will see this in loops.
train, test = next(islice(cv.split(X, y), 2, 3))cv is from sklearn.
Thanks for the info. That makes sense. When a line is executed multiple times, only the maximum is displayed, which might lead to some confusing information being displayed. PRs or ideas on how to handle this better are welcome.
Unless we find other cases where this happens, I guess this is a duplicate of #9.
@fabianp proposed a fix for this in #202. @karan10111 @alowayed @nathantsoi could you please try
pip install https://github.com/fabianp/memory_profiler/archive/fix_195.zipand rerun your examples?Reacted by Eric Fiske and Harsh Gupta- added a commit that references this issue
on Aug 16, 2018
i am using memory profile as follows:
however the results seem strange, I see large negative increments, much larger than any calls with positive increments: