update cudnn (#2265)

doc/fluid/install/index_cn.rst

@@ -32,7 +32,7 @@
 =================================
 
 * 目前 **PaddlePaddle** 仅支持 **NVIDIA** 显卡的 **CUDA** 驱动
-* 需要安装 `cuDNN <https://docs.nvidia.com/deeplearning/sdk/cudnn-install/>`_ ，版本要求 7.3+(For CUDA9/10)
+* 需要安装 `cuDNN <https://docs.nvidia.com/deeplearning/sdk/cudnn-install/>`_ ，版本要求 7.6+(For CUDA9/10)
 * 如果您需要 GPU 多卡模式，需要安装 `NCCL 2 <https://developer.nvidia.com/nccl/>`_
 
     * 仅 Ubuntu/CentOS 支持 NCCL 2 技术

doc/fluid/install/index_en.rst

@@ -32,7 +32,7 @@ The manuals will guide you to build and install PaddlePaddle on your 64-bit desk
 =================================
 
 * Currently, **PaddlePaddle** only supports **CUDA** driver of **NVIDIA** graphics card.
-* You need to install `cuDNN <https://docs.nvidia.com/deeplearning/sdk/cudnn-install/>`_ , and version 7.3+ is required(For CUDA9/10) 
+* You need to install `cuDNN <https://docs.nvidia.com/deeplearning/sdk/cudnn-install/>`_ , and version 7.6+ is required(For CUDA9/10) 
 * If you need GPU multi-card mode, you need to install `NCCL 2 <https://developer.nvidia.com/nccl/>`_
 
     * Only Ubuntu/CentOS support NCCL 2

doc/fluid/install/install_CentOS.md

@@ -64,8 +64,8 @@
 
 * 如果您的计算机有NVIDIA® GPU，请确保满足以下条件并且安装GPU版PaddlePaddle
 
-	* **CUDA 工具包10.0配合cuDNN v7.3+(如需多卡支持，需配合NCCL2.3.7及更高)**
-	* **CUDA 工具包9.0配合cuDNN v7.3+(如需多卡支持，需配合NCCL2.3.7及更高)**
+	* **CUDA 工具包10.0配合cuDNN v7.6+(如需多卡支持，需配合NCCL2.3.7及更高)**
+	* **CUDA 工具包9.0配合cuDNN v7.6+(如需多卡支持，需配合NCCL2.3.7及更高)**
 	* **GPU运算能力超过1.0的硬件设备**
 
 		您可参考NVIDIA官方文档了解CUDA和CUDNN的安装流程和配置方法，请见[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/)，[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/)

doc/fluid/install/install_CentOS_en.md

@@ -65,19 +65,19 @@
 
 * If your computer has NVIDIA® GPU, please make sure that the following conditions are met and install the GPU version of PaddlePaddle
 
-	* **CUDA toolkit 10.0 with cuDNN v7.3+(for multi card support, NCCL2.3.7 or higher)**
-	* **CUDA toolkit 9.0 with cuDNN v7.3+(for multi card support, NCCL2.3.7 or higher)**
-	* **Hardware devices with GPU computing power over 1.0**
+    * **CUDA toolkit 10.0 with cuDNN v7.6+(for multi card support, NCCL2.3.7 or higher)**
+    * **CUDA toolkit 9.0 with cuDNN v7.6+(for multi card support, NCCL2.3.7 or higher)**
+    * **Hardware devices with GPU computing power over 1.0**
 
-		You can refer to NVIDIA official documents for installation process and configuration method of CUDA and cudnn. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/)，[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/)
+        You can refer to NVIDIA official documents for installation process and configuration method of CUDA and cudnn. Please refer to [CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/)，[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/)
 
 * 如果您需要使用多卡环境请确保您已经正确安装nccl2，或者按照以下指令安装nccl2（这里提供的是CentOS 7，CUDA9，cuDNN7下nccl2的安装指令），更多版本的安装信息请参考NVIDIA[官方网站](https://developer.nvidia.com/nccl):
 
 
-		wget http://developer.download.nvidia.com/compute/machine-learning/repos/rhel7/x86_64/nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm
-		rpm -i nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm
-		yum update -y
-		yum install -y libnccl-2.3.7-2+cuda9.0 libnccl-devel-2.3.7-2+cuda9.0 libnccl-static-2.3.7-2+cuda9.0
+        wget http://developer.download.nvidia.com/compute/machine-learning/repos/rhel7/x86_64/nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm
+        rpm -i nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm
+        yum update -y
+        yum install -y libnccl-2.3.7-2+cuda9.0 libnccl-devel-2.3.7-2+cuda9.0 libnccl-static-2.3.7-2+cuda9.0
 
 ## Installation method
 
@@ -101,12 +101,12 @@ Here is pip installation
 You can[Verify installation succeeded or not](#check)，if you have any questions, you can refer to [FAQ](./FAQ.html)
 
 
-Note: 
+Note:
 
 * If it is python2.7, it is recommended to use the `python` command; if it is python3.x, it is recommended to use the 'python3' command
 
 
-* `python -m pip install paddlepaddle-gpu==2.0.0a0 -i https://pypi.tuna.tsinghua.edu.cn/simple` This command will install the PaddlePaddle that supports CUDA 10.0 cuDNN v7. 
+* `python -m pip install paddlepaddle-gpu==2.0.0a0 -i https://pypi.tuna.tsinghua.edu.cn/simple` This command will install the PaddlePaddle that supports CUDA 10.0 cuDNN v7.
 
 
 * Download the latest stable installation package by default. For development installation package, please refer to [here](./Tables.html#ciwhls)

doc/fluid/install/install_Ubuntu.md

@@ -64,8 +64,8 @@
 * 如果您的计算机没有 NVIDIA® GPU，请安装CPU版的PaddlePaddle
 
 * 如果您的计算机有 NVIDIA® GPU，并且满足以下条件，推荐安装GPU版的PaddlePaddle
-	* **CUDA 工具包10.0配合cuDNN v7.3+(如需多卡支持，需配合NCCL2.3.7及更高)**
-	* **CUDA 工具包9.0配合cuDNN v7.3+(如需多卡支持，需配合NCCL2.3.7及更高)**
+	* **CUDA 工具包10.0配合cuDNN v7.6+(如需多卡支持，需配合NCCL2.3.7及更高)**
+	* **CUDA 工具包9.0配合cuDNN v7.6+(如需多卡支持，需配合NCCL2.3.7及更高)**
 	* **GPU运算能力超过1.0的硬件设备**
 
 

doc/fluid/install/install_Ubuntu_en.md

@@ -64,19 +64,19 @@
 * If your computer doesn't have NVIDIA® GPU, please install CPU version of PaddlePaddle
 
 * If your computer has NVIDIA® GPU, and meet the following conditions, we command you to install PaddlePaddle
-	* **CUDA toolkit 10.0 with cuDNN v7.3+(for multi card support, NCCL2.3.7 or higher)**
-	* **CUDA toolkit 9.0 with cuDNN v7.3+(for multi card support, NCCL2.3.7 or higher)**
-	* **Hardware devices with GPU computing power over 1.0**
+    * **CUDA toolkit 10.0 with cuDNN v7.6+(for multi card support, NCCL2.3.7 or higher)**
+    * **CUDA toolkit 9.0 with cuDNN v7.6+(for multi card support, NCCL2.3.7 or higher)**
+    * **Hardware devices with GPU computing power over 1.0**
 
 
-	You can refer to NVIDIA official documents for installation process and configuration method of CUDA and cudnn. Please refer to[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/)，[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/)
+    You can refer to NVIDIA official documents for installation process and configuration method of CUDA and cudnn. Please refer to[CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/)，[cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/)
 
 * If you need to use multi card environment, please make sure that you have installed nccl2 correctly, or install nccl2 according to the following instructions (here is the installation instructions of nccl2 under ubuntu 16.04, CUDA9 and cuDNN7). For more version of installation information, please refer to NVIDIA[official website](https://developer.nvidia.com/nccl):
 
 
-	wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb
-	dpkg -i nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb
-	sudo apt-get install -y libnccl2=2.3.7-1+cuda9.0 libnccl-dev=2.3.7-1+cuda9.0
+    wget https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb
+    dpkg -i nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb
+    sudo apt-get install -y libnccl2=2.3.7-1+cuda9.0 libnccl-dev=2.3.7-1+cuda9.0