POPs gpu memory optimization

Hello there!

I have a big master project setup for VJing in which i store many networks in a Base COMPs and using the “allowcooking” approach to optimize the resources so the cooking is happening only for the selected BaseCOMPs. Altough, the Base COMPs which comtain networks with POP operators inside seems to still consume the gpu memory nonetheless that the Base COMP’s cooking is deactivated.

So eventually my system runs out of memory when i switch between several POPs networks, because each of them continues to process.

I used Probe from the pallet to figure this out.

Have anyone encountered this issue and could you share any solutions to this problem?

P. S.

The FreeExtraGpuMemory button doesn’t help, because it frees only the extra memory used.

Thank you!

Hello @ComplexCarrot,

you can use the .progressiveUnload() method to unload a component, it worked as expected with a simple test case.

We’ll make the behavior for POPs consistent with TOPs (disabling cooking on the comp frees up GPU memory).

Best,
Vincent