Silver Clear Model Cache
The Reset Button for Silver's Shared Model Cache
This node does one thing, and it only matters if you use the other two Silver cache loaders. It's a button that empties the shared model cache those nodes fill - click it (or wire it in) and every cached diffusion model and text encoder is dropped so the RAM comes back.
What it is
Silver Clear Model Cache is an output node with no inputs and no outputs. Run it and it wipes the module-level cache that Silver Shared Diffusion Model and Silver Shared Text Encoder build up, then reports how many entries it cleared. That's the entire job description. If you aren't using the shared loaders, this node does literally nothing - there's no cache for it to touch.
Why you'd reach for it
The shared loaders work by pinning model weights in RAM so a queued server doesn't re-read an 8 GB file from disk on every prompt. That's a great trade when you're reusing the same model a hundred times, and a slow leak when you're not. If you switch models between jobs - or a long batch run has accumulated several big encoders in memory and you want the RAM back before the next job - a clear is the clean reset. Drop it at the end of a workflow or trigger it from an API call when you know the next job needs different weights.
It's also the cheap answer to a specific failure mode: because the cache is keyed and reused, a stale cached model is exactly the kind of "silently wrong" bug that hides. If you ever suspect you're generating with the wrong weights, hit clear and reload fresh.
How it works
Under the hood it's two dicts - the cache itself and a per-key hit counter - and the node just empties both. The hit counter matters because the shared loaders use it in their change-detection to decide whether a run should reuse the cached model. Clearing it means the next prompt genuinely reloads from disk, no shortcuts.
One thing it does not do: touch ComfyUI's normal model management. This only clears Silver's own pinned cache. If you're chasing VRAM and expecting it to behave like ComfyUI's free or unload nodes, that's not this node. It frees RAM; ComfyUI's own memory management still handles VRAM.
Installing it
Silver Nodes is a small utility pack by SilverAndJade, and this node ships in it - there's nothing separate to install. ComfyUI Manager: search "ComfyUI Silver Nodes" and install, or:
cd ComfyUI/custom_nodes
git clone https://github.com/SilverAndJade/comfyui-silver-nodes
Then restart ComfyUI. No models, no dependencies beyond the pack's own requirements.txt (which Manager installs for you).
Common issues
The only real trap is expecting it to be something it isn't. It's a RAM release, not a VRAM management tool, and it only affects Silver's cache. The bigger thing to know is that the cache nodes it pairs with want a manual patch to ComfyUI's comfy/model_management.py to fully deduplicate loaded weights - if you skipped that patch, clearing the cache can't fix the doubled VRAM, because the doubling happens outside Silver's cache entirely. Clear is a maintenance tool, not a patch substitute.
Inputs (0)
No inputs
Outputs (0)
No outputs