diff --git a/docs/requirements.txt b/docs/requirements.txt
index c7d94486..cfa9c93a 100644
--- a/docs/requirements.txt
+++ b/docs/requirements.txt
@@ -1,8 +1,4 @@
-numpy>=1.14.2
-http://download.pytorch.org/whl/cpu/torch-0.4.1-cp36-cp36m-linux_x86_64.whl
-torchvision>=0.1.8
-sphinx-rtd-theme==0.4.1
-tensorboardX>=1.4
-tqdm>=4.28.1
-ipython>=6.4.0
-ipython-genutils>=0.2.0
\ No newline at end of file
+sphinx==3.2.1
+docutils==0.16
+sphinx-rtd-theme==0.5.0
+readthedocs-sphinx-search==0.1.0rc3
\ No newline at end of file
diff --git a/docs/source/tutorials/文本分类.rst b/docs/source/tutorials/文本分类.rst
index 73686916..30f6cf4f 100644
--- a/docs/source/tutorials/文本分类.rst
+++ b/docs/source/tutorials/文本分类.rst
@@ -291,7 +291,7 @@ fastNLP提供了Trainer对象来组织训练过程，包括完成loss计算(所
 
 
 PS: 使用Bert进行文本分类
-~~~~~~~~~~~~~~~~~~~~
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 .. code-block:: python
 
@@ -368,7 +368,7 @@ PS: 使用Bert进行文本分类
 
 
 PS: 基于词进行文本分类
-~~~~~~~~~~~~~~~~~~~~
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 由于汉字中没有显示的字与字的边界，一般需要通过分词器先将句子进行分词操作。
 下面的例子演示了如何不基于fastNLP已有的数据读取、预处理代码进行文本分类。
diff --git a/readthedocs.yml b/readthedocs.yml
index e6d5bafd..73935f9a 100644
--- a/readthedocs.yml
+++ b/readthedocs.yml
@@ -7,10 +7,11 @@ build:
   image: latest
 
 python:
-  version: 3.6
+  version: 3.7
   install:
+    - requirements: docs/requirements.txt
     - method: setuptools
       path: .
 
 formats:
-  - htmlzip
\ No newline at end of file
+  - htmlzip
diff --git a/setup.py b/setup.py
index 714a7e42..1ebd4b7f 100644
--- a/setup.py
+++ b/setup.py
@@ -23,7 +23,7 @@ setup(
     long_description_content_type='text/markdown',
     license='Apache License',
     author='Fudan FastNLP Team',
-    python_requires='>=3.6',
+    python_requires='>=3.7',
     packages=pkgs,
     install_requires=reqs.strip().split('\n'),
 )