From 09a22c75dc622d199ed35102a12f3ac88f10cef9 Mon Sep 17 00:00:00 2001
From: secretzer0 <secretzer0@vyzon.ai>
Date: Sat, 15 Aug 2026 22:07:08 -0500
Subject: [PATCH] Fix GPU memory leak in voice conditioning cache

Wrap generation in torch.inference_mode so cached conds do not retain autograd graphs (~200 MB per voice).
---
 engine.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/engine.py b/engine.py
index a2be1e9..1a4407b 100644
--- a/engine.py
+++ b/engine.py
@@ -487,7 +487,7 @@ def synthesize(
         # Call the core model's generate method.
         # autocast promotes float32 inputs to bfloat16 to match T3/S3Gen weights,
         # keeping numerically sensitive ops (softmax, norms) in float32 automatically.
-        with torch.autocast("cuda", dtype=torch.bfloat16, enabled=BF16_ENABLED):
+        with torch.inference_mode(), torch.autocast("cuda", dtype=torch.bfloat16, enabled=BF16_ENABLED):
             if loaded_model_type == "multilingual":
                 wav_tensor = chatterbox_model.generate(
                     text=text,