diff --git a/README.md b/README.md
index 009671b..b70c14a 100644
--- a/README.md
+++ b/README.md
@@ -197,7 +197,7 @@ device_map = infer_auto_device_map(model, max_memory={0: "6GiB", "cpu": "128GiB"
 
 One with A100 might try to set 38Gb to a GPU0 and try to inference the model completely in the GPU VRAM.
 
-One with 4*A100 might wish to use: 0: "38GiB", 1: "38Gb" etc.
+One with 4*A100 might wish to use: {0: "38GiB", 1: "38GiB", 2: "38GiB", 3: "38GiB", "cpu":"128GiB"}.
 
 For me, with 6Gb for 3070ti, this works three times slower against pure CPU inference.