You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
My movie is 1129 frames of 3000x3000 images. It has about ~450 cells per frame with a big drop in cell number towards the end. Do you all have any suggestions on how to diagnose the kernel dying event? The kernel doesn't die with a lower max_distance, but the cells aren't properly linked at the start of the movie when they are moving quickly.
The text was updated successfully, but these errors were encountered:
Your problem is taking a long time to converge, so I would change the solution_gap to 0.005, the algorithm might be using more memory storing intermediate solutions to reach a global optimal.
These last few percents of the gap might change a handful associations and are often not critical.
I also suggest checking if you really benefit from the high number of parallel threads during tracking (config.tracking_config.n_threads).
Each thread in gurobi explores an instances in parallel, so you memory usage scale with the number of threads.
ILPs are a bit weird, sometimes more threads aren't very beneficial.
I have a 128GB of memory on my workstation and usage was hovering around 60% when I checked but I didn't track it carefully. I'll try with your suggestions and report back. Thanks!
I have
ultrack
running successfully on my machine, but I keep getting kernel dying errors during the Gurobi optimization section.Output from track function:
Config:
My movie is 1129 frames of 3000x3000 images. It has about ~450 cells per frame with a big drop in cell number towards the end. Do you all have any suggestions on how to diagnose the kernel dying event? The kernel doesn't die with a lower
max_distance
, but the cells aren't properly linked at the start of the movie when they are moving quickly.The text was updated successfully, but these errors were encountered: