From ec24e55f6a9798dd281f5856158b0edf384d3a70 Mon Sep 17 00:00:00 2001
From: shumingma <shumma@microsoft.com>
Date: Wed, 23 Nov 2022 20:25:12 -0800
Subject: [PATCH] update pic path

---
 README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 9e5c774..8a28d1d 100644
--- a/README.md
+++ b/README.md
@@ -93,7 +93,7 @@ We plan to provide more examples regarding different tasks (e.g. vision pretrain
 ### Stability Evaluation
 
 <p align="center">
-  <img src="./assets/convergence.png" width="800"/>
+  <img src="https://publicmodel.blob.core.windows.net/torchscale/pic/convergence.png" width="800"/>
 </p>
 
 The training curve is smooth by using TorchScale, while the baseline Transformer cannot converge.
@@ -101,7 +101,7 @@ The training curve is smooth by using TorchScale, while the baseline Transformer
 ### Scaling-up Experiments
 
 <p align="center">
-  <img src="./assets/scaling_curve.png" width="800"/>
+  <img src="https://publicmodel.blob.core.windows.net/torchscale/pic/scaling_curve.png" width="800"/>
 </p>
 
 TorchScale supports arbitrary depths and widths, successfully scaling-up the models without pain.