22

app.py

@@ -58,6 +58,17 @@ pipe_lora = QwenImagePipeline.from_pretrained(
     vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 0.5,
 )
 
+vram_config = {
+    "offload_dtype": "disk",
+    "offload_device": "disk",
+    "onload_dtype": torch.bfloat16,
+    "onload_device": "cpu",
+    "preparing_dtype": torch.bfloat16,
+    "preparing_device": "cuda",
+    "computation_dtype": torch.bfloat16,
+    "computation_device": "cuda",
+}
+
 pipe_imagen = QwenImagePipeline.from_pretrained(
     torch_dtype=torch.bfloat16,
     device="cuda",