Don’t repeat my mistake with the ‘Custom model folder’

[jaydeel]

This may be common knowledge, but it tripped me up this week.

Over a week ago I was experimenting with six custom models. I had them all loaded in the ‘Custom model folder’ (C:\DeepStack\MyModels). Eventually I settled on using only three of the six models.

Here’s where I made my mistake. I failed to remove the unused models (*.pt files) from the folder. For the past week I started noticing more CPU usage warnings and high RAM usage warnings.

Upon investigating I relearned that DeepStack creates a python.exe instance for each *.pt model in the ‘Custom model folder’ - regardless of whether the models are actually used by a Blue Iris camera.
.
As I had settled on only three custom models, I was allowing DeepStack to use about 5GB of unnecessary RAM to my 16GB system (~1.5GB ‘working set’ RAM per model). In total, my system was using 14/16GB (87% as suggested by the Android app screenshot). After moving the files, I was using 10.7/16GB (66%).

See the attached screenshots for the details.

BEFORE…


AFTER…

 

[TBurt]

This may be common knowledge, but it tripped me up this week.

Over a week ago I was experimenting with six custom models. I had them all loaded in the ‘Custom model folder’ (C:\DeepStack\MyModels). Eventually I settled on using only three of the six models.

Here’s where I made my mistake. I failed to remove the unused models (*.pt files) from the folder. For the past week I started noticing more CPU usage warnings and high RAM usage warnings.

Upon investigating I relearned that DeepStack creates a python.exe instance for each *.pt model in the ‘Custom model folder’ - regardless of whether the models are actually used by a Blue Iris camera.
.
As I had settled on only three custom models, I was allowing DeepStack to use about 5GB of unnecessary RAM to my 16GB system (~1.5GB ‘working set’ RAM per model). In total, my system was using 14/16GB (87% as suggested by the Android app screenshot). After moving the files, I was using 10.7/16GB (66%).

See the attached screenshots for the details.

BEFORE…
View attachment 133659 View attachment 133660 View attachment 133661

AFTER…
View attachment 133662 View attachment 133664 View attachment 133668

I did the same thing and found out the hard way. My computer's memory usage was hitting 97%-99% and getting the BI error Ai not responding. Then BI would stop responding for several seconds, then start working again only to freeze up again. I now have two Deep Stack model folders. One with all my custom models, and one that has only the ones I am using right now. I meant to say something on here about it but got busy and forgot. Thanks for the heads up. Maybe you will save someone else having to troubleshoot their Blue Iris server/system.

 

[tech101]

Aha, Thank you so much for this. I Think I have some sitting which are not being utilized as well never knew just by having it there can use unnecessary resources. Thank you so much for sharing this.

 

[Sybertiger]

Thanks, I was wondering why I had three Python processes running at the same time continually.