Trying to fix CUDA OOM in Fooocus.

[1990Billsfan]

Does anyone know where to put commandline arguments in a Fooocus installation?

Wanting to use :

" set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128"

Which I found in a post about a user having Cuda OOM errors in AUTOMATIC 1111...

Would I use this same argument for Fooocus? And if so where would it go?

Running Fooocus with 32 GB RAM and GTX 1650

My 1st Generation is very slow (as expected) but completes properly, but if I try to generate another pic I get the dreaded CUDA OOM crash accompanied by this message:

"If reserved but unallocated memory is large try setting max_split_size"

My "reserved but unallocated memory" IS large (over 1.5 GB) so I am trying to follow the software's advice and run the argument it's asking for, just not sure where it would go.

Any help someone could provide would be greatly appreciated.