diff --git a/research/cv/vit_base/README.md b/research/cv/vit_base/README.md
index 2034d182021a8c4ffc15e90ed08cf7ec39ff42bd..1750f865a115cf01939310722b498b2fb77c6bb7 100644
--- a/research/cv/vit_base/README.md
+++ b/research/cv/vit_base/README.md
@@ -70,8 +70,17 @@ Adopt [mixed precision](https://www.mindspore.cn/docs/programming_guide/zh-CN/r1
 
 # Quick start
 
-After installing MindSpore through the official website，you can follow the steps below for training and evaluation，in particular，before training, you need to download the official baseImageNet21k pre-trained model [ViT-B_16](https://console.cloud.google.com/storage/vit_models/) , and convert it to the ckpt format model supported by MindSpore，named "cifar10_pre_checkpoint_based_imagenet21k.ckpt"，place the training set and test set data in the same level directory:
-Note: you can use .npz file, but only for ViT-base-16. For that, type the path to .npz in `config.py`.
+After installing MindSpore through the official website，you can follow the steps below for training and evaluation，in particular，before training, you need to download the official base [ImageNet21k](https://console.cloud.google.com/storage/vit_models/) pre-trained model [ViT-B_16](http://storage.googleapis.com/vit_models/imagenet21k/ViT-B_16.npz) , and convert it to the ckpt format model supported by MindSpore，named "cifar10_pre_checkpoint_based_imagenet21k.ckpt"，place the training set and test set data in the same level directory:
+
+ ```text
+└─dataset
+    ├─cifar10
+        ├─cifar-10-batches-bin
+        └─cifar-10-verify-bin
+    └─cifar10_pre_checkpoint_based_imagenet21k.ckpt
+```
+
+Note: you can use .npz file, but only for ViT-base-16. For that, change the `checkpoint_path` parameter in `config.py` to the path of .npz file.
 
 - Ascend or GPU
 
diff --git a/research/cv/vit_base/README_CN.md b/research/cv/vit_base/README_CN.md
index 0e303205e6d7d4e55a3c1d13ae1f4a756ed00032..0f5002833cfe0ac363d57c1832b021994726609b 100644
--- a/research/cv/vit_base/README_CN.md
+++ b/research/cv/vit_base/README_CN.md
@@ -70,7 +70,17 @@ vit_base的总体网络架构如下： [链接](https://arxiv.org/abs/2010.11929
 
 # 快速入门
 
-通过官方网站安装MindSpore后，您可以按照如下步骤进行训练和评估，特别地，进行训练前需要先下载官方基于ImageNet21k的预训练模型[ViT-B_16](https://console.cloud.google.com/storage/vit_models/) ，并将其转换为MindSpore支持的ckpt格式模型，命名为"cifar10_pre_checkpoint_based_imagenet21k.ckpt"，和训练集测试集数据放于同一级目录下：
+通过官方网站安装MindSpore后，您可以按照如下步骤进行训练和评估，特别地，进行训练前需要先下载官方基于[ImageNet21k](https://console.cloud.google.com/storage/vit_models/)的预训练模型[ViT-B_16](http://storage.googleapis.com/vit_models/imagenet21k/ViT-B_16.npz) ，并将其转换为MindSpore支持的ckpt格式模型，命名为"cifar10_pre_checkpoint_based_imagenet21k.ckpt"，和训练集测试集数据放于同一级目录下：
+
+ ```text
+└─dataset
+    ├─cifar10
+        ├─cifar-10-batches-bin
+        └─cifar-10-verify-bin
+    └─cifar10_pre_checkpoint_based_imagenet21k.ckpt
+```
+
+注：你可以使用 .npz 格式的文件，但是仅适用于ViT-base-16。为此，请将`config.py`文件的`checkpoint_path`参数改为 .npz 文件的路径。
 
 - Ascend or GPU 处理器环境运行