From d4f895042d3601fd053956c917dbff176c84694d Mon Sep 17 00:00:00 2001
From: Rattus <rattus128@gmail.com>
Date: Tue, 20 Jan 2026 21:57:28 +1000
Subject: [PATCH] execution: remove per node gc.collect()

This isn't worth it and the likelyhood of inference leaving a complex
data-structure with cyclic reference behind is now. Remove it.

We would replace it with a condition on nodes that actually touch the
GPU which might be win.
---
 execution.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/execution.py b/execution.py
index 197e13f38..4148baaf5 100644
--- a/execution.py
+++ b/execution.py
@@ -530,7 +530,6 @@ async def execute(server, dynprompt, caches, current_item, extra_data, executed,
                 #FIXME: this is probably a little zealous
                 # Torch code comments says some stuff about not actually freeing tensors on mempool
                 #context release. Explicitly garbage collect now.
-                gc.collect()
                 torch.cuda.empty_cache()
 
             if has_pending_tasks: