From 177c6c158bd1634ab91b8bafda4b77eb07fa92cd Mon Sep 17 00:00:00 2001
From: LastWhisper <pkuwkl@gmail.com>
Date: Thu, 30 May 2024 00:32:58 +0800
Subject: [PATCH] Update README.md, add a header figure.

---
 README.md | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 6773dbb..958346b 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,10 @@
 
 > Google DeepMind: Mixture-of-Depths Unofficial Implementation.
 
+<div align="center">
+<img width="624" alt="MoD" src="https://github.com/Mixture-AI/Mixture-of-Depths/assets/103916249/4e84ca8b-4ace-422e-a8e9-84070a82b14e">
+</div>
+
 ## TODO List
 
 - [ ] Enable the **batching forward** operation.
@@ -25,4 +29,4 @@ This section informs us how to solve the non-causal problem of the top-k operati
 
 > Section 5. "If a token does not participate in self-attention at a certain block, then later tokens will also not be able to attend to it."
 
-This section tells us that for any Transformer block, whether a token participates in computation is determined after the first routing and will not change afterward. (✅ implement)
\ No newline at end of file
+This section tells us that for any Transformer block, whether a token participates in computation is determined after the first routing and will not change afterward. (✅ implement)