xinyun/ComfyUI
1
0
Fork 0
mirror of https://git.datalinker.icu/comfyanonymous/ComfyUI synced 2026-08-14 10:56:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix memory leak by properly detaching model finalizer

Browse Source 
When unloading models in load_models_gpu(), the model finalizer was not
being explicitly detached, leading to a memory leak. This caused
linear memory consumption increase over time as models are repeatedly
loaded and unloaded.

This change prevents orphaned finalizer references from accumulating in
memory during model switching operations.
This commit is contained in:
guy 2025-09-22 05:01:45 +00:00
parent 27bc181c49
commit 5a0ab8f214
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
comfy/model_management.py
View File

@ -645,7 +645,9 @@ def load_models_gpu(models, memory_required=0, force_patch_weights=False, minimu
if loaded_model.model.is_clone(current_loaded_models[i].model): if loaded_model.model.is_clone(current_loaded_models[i].model):
to_unload = [i] + to_unload to_unload = [i] + to_unload
for i in to_unload: for i in to_unload:
current_loaded_models.pop(i).model.detach(unpatch_all=False) model_to_unload = current_loaded_models.pop(i)
model_to_unload.model.detach(unpatch_all=False)
model_to_unload.model_finalizer.detach()
total_memory_required = {} total_memory_required = {}
for loaded_model in models_to_load: for loaded_model in models_to_load: