Update app.py

app.py

@@ -47,6 +47,13 @@ SLIDER_MIN_W, SLIDER_MAX_W = 128, 1280
 print("Initializing WanTI2V pipeline...")
 device = "cuda" if torch.cuda.is_available() else "cpu"
 device_id = 0 if torch.cuda.is_available() else -1
+# Set CUDA memory allocation configuration
+os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
+
+# Clear CUDA cache before model initialization
+if torch.cuda.is_available():
+    torch.cuda.empty_cache()
+    
 pipeline = wan.WanTI2V(
     config=cfg,
     checkpoint_dir=ckpt_dir,
@@ -55,8 +62,8 @@ pipeline = wan.WanTI2V(
     t5_fsdp=False,
     dit_fsdp=False,
     use_sp=False,
-    t5_cpu=False,
-    init_on_cpu=False,
+    t5_cpu=True,  # Use CPU for T5 model
+    init_on_cpu=True,  # Initialize on CPU first
     convert_model_dtype=True,
 )
 print("Pipeline initialized and ready.")