Better switching between cache none and not for cleaing models in workflows?

### Feature Idea

A better way of implementing --cache-none or its function?

I have found using --cache-none very helpful for using LTX 2.3 on my peasant system, it dumps the gemma model before loading the main model reducing overload on my system. The issue is when I am using image models that don't require this, every generation the gguf clip and so on is getting dumped from the memory slowing everything down. So far I am having to reset comfyui each time and take the argument in and out as needed.

Is there or can there be a better way of switching?

I've tried RAM/VRAM/Cache cleaning nodes between my gemma model and the sampler in my LTX workflow but nothing seems to dump the model or clean things out the same way --cache-none does. Unless I am routing it wrong.

I've tried nodes that keep models in memory too but of course this isn't going to work is it as --cache-none overrrides all that?

Can anyone understand the inconvenience here and knows if there's a solution better than having to toggle the argument on and off and restart each time.

Thanks.

### Existing Solutions

_No response_

### Other

_No response_