diff --git a/README.md b/README.md
index 5272bd9..223dc87 100644
--- a/README.md
+++ b/README.md
@@ -176,7 +176,7 @@ If you encounter a **RuntimeError** regarding "Inference Tensors":
 * Ensure you are using a **Batch Size** large enough to split.
 * The node uses a "Deep Detach" strategy (`.detach().clone()`) to satisfy the version counter requirements of the KSampler.
 * if you see the message "RuntimeError: Expected all tensors to be on the same device, but got mat1 is on cuda:0, different from other tensors on cuda:1 (when checking argument in method wrapper_CUDA_addmm)
-" after changing the percentage of the GPU'S for the second run. restart comfyui
+" after changing the percentage of the GPU'S for the second run. restart comfyui or add **--cache-none** in the ComfyUI startup arguments
 
 ### Slower than single GPU
 #### Common causes: