im not sure what the issue is but these nodes are GIVING me memory errors.

[RealRebelAI]

i ran a series of testss for these nodes, grouped with the dynamic ramcache node. inside an fp8 Z-Image-Turbo workflow (the least vram intense model possible) and now every other image is giving Out Of Memory errors. the workflow worked flawlessly before i introduced the nodes. ran on autoregressive and non. just errors. i think they place weight onto the vram just being used in the owrkflow. im not sure how to explain it any better as im so new to this world but i will say the workflow ran better without the clutter of nodes.

[Windecay]

i ran a series of testss for these nodes, grouped with the dynamic ramcache node. inside an fp8 Z-Image-Turbo workflow (the least vram intense model possible) and now every other image is giving Out Of Memory errors. the workflow worked flawlessly before i introduced the nodes. ran on autoregressive and non. just errors. i think they place weight onto the vram just being used in the owrkflow. im not sure how to explain it any better as im so new to this world but i will say the workflow ran better without the clutter of nodes.

ram_pressure is a newly added parameter in comfyUI, which I consider to be an experimental feature. This node only provides a switch, with the actual implementation relying on parameters within the core, leading to many potential issues in use. I believe this part of comfy requires more testing.