From 0e3c42a830ef87941983a6a494e8df0f90e616d2 Mon Sep 17 00:00:00 2001
From: Adrian Corduneanu <adrian.dc@gmail.com>
Date: Wed, 26 Feb 2025 02:56:57 -0800
Subject: [PATCH] Update text2video.py to reduce GPU memory by emptying cache
 (#44)

* Update text2video.py to reduce GPU memory by emptying cache

If offload_model is set, empty_cache() must be called after the model is moved to CPU to actually free the GPU. I verified on a RTX 4090 that without calling empty_cache the model remains in memory and the subsequent vae decoding never finishes.

* Update text2video.py only one empty_cache needed before vae decode
---
 wan/text2video.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/wan/text2video.py b/wan/text2video.py
index 96cfa78..2400545 100644
--- a/wan/text2video.py
+++ b/wan/text2video.py
@@ -252,6 +252,7 @@ class WanT2V:
             x0 = latents
             if offload_model:
                 self.model.cpu()
+                torch.cuda.empty_cache()
             if self.rank == 0:
                 videos = self.vae.decode(x0)