Using RTX 4090, but yet I end up with 'The application was unable to allocate enough GPU memory to use this model.'

[ReneRu]

My system is capable of doing pretty much anything, from Stable Diffusion to other generative AI stuff, but one thing it cannot seem to do is use three different models at the same time in UVR5 - MDX23C-InstaVoc HQ; UVR-MDX-NET Inst HQ4 and Demucs v4 htdemucs_ft. Do you have any ideas on how to bypass that stuff? Is there a reliable workaround?

When looking at the task manager, my GPU is utilized between 98-100%, but only 6-7GB out of 24GB of VRAM and 0.2GB out of 32GB of Shared GPU memory are used during the process. The same happens when I run the Stable diffusion client locally, but it never makes the operation crash/stop because of that.

Edit:
When I remove Demucs and only keep the other two, it finalizes the render. But I'd really want to use the Demucs as well.