TL;DR: ComfyUI users are experiencing significant generation slowdowns due to an open bug causing models to reload from disk instead of RAM.
Summary: Multiple open issues on ComfyUI's GitHub indicate a bug where models are repeatedly reloaded from disk rather than being cached in RAM. This behavior drastically increases generation times, with some users reporting a 30-40% increase in processing duration.
Why it matters: AI developers using ComfyUI should be aware of this performance degradation and monitor the linked GitHub issues for updates and potential workarounds. This highlights the importance of efficient model management in local AI inference tools.
Source: reddit