1hew-ComfyUI-1hewNodes/web/docs/1hew_MemoryCleanup/en.md

Memory Cleanup - Memory Cleanup

Node Purpose: Memory Cleanup requests ComfyUI to release execution cache and model memory after the current task completes, ideal for cleaning up RAM/VRAM between queued executions.

Inputs

Name	Port	Type	Default	Range	Description
anything	optional	*	-	-	Arbitrary input, passed through to output.
unload_model	-	BOOLEAN	False	True / False	Whether to request unloading loaded models after the task. Saves VRAM/RAM but may require model reloading on the next run.

Outputs

Name	Type	Description
output	*	Passes through the anything input value.

Features

• Calls gc.collect() to trigger Python garbage collection on execution.
• Invokes ComfyUI's soft_empty_cache to free PyTorch cache.
• Sets the free_memory flag via prompt_queue to request execution cache release.
• When unload_model=True, additionally sets unload_models flag to unload loaded models.

Typical Usage

• Insert this node between queued tasks to prevent VRAM from accumulating.
• Enable unload_model=True when switching models or workflows to release old models.
• Place at the end of a workflow to auto-cleanup after each execution.

Notes & Tips

• This is an output node (is_output_node=True), so ComfyUI ensures it executes at the end of the workflow.
• Enabling unload_model causes model reload on the next run, increasing single-run time; use as needed.
• The anything input accepts any type, allowing you to chain preceding nodes to ensure cleanup triggers at the correct execution order.