Master information platform : the basis of today learn about Tensorflow2.0
Core library, @ tf.function, can easily be dynamic graphic language, become a static map, calculated in a way to accelerate
TensorFlow Lite
TensorFlow.JS
TensorFlow Extended
Constitute TensorFlow ecosystem
Advantage:
1, GPU acceleration
When large amounts of data reflected in the calculation, the computation time. If the CPU operation, the calculations are completed by serial mode
GPU will speed operation, parallel operation, fast run.
2, automatic derivative
It comes with automatic derivation tool to facilitate rapid derivation.
3, neural network
Direct call interface TensorFlow provided, we do not need to own to achieve.
tf.matmul layers.Demse
tf.nn.conv2d layers.Conv2D
tf.nn.relu layers.SimpleRNN
tf.nn.max_pool2d layers.LSTM
tf.nn.sigmoid layers.RelU
tf.nn.softmax layers.MaxPool2D
First, the installation environment
Win10
Anaconda 、Python3.7
CUDA 10.0 cuDNN
TensorFlow 2.0
PyCharm
1, Anaconda installation, not repeat them here, not the friends themselves Baidu
Tests are not installed Anaconda, directly open the command line, type: conda list
Contents shown below will appear, i.e. installation
2, CUDA installation
Before installation, assuming your PC is Nvidia graphics card, the graphics cards GTX 1060 6g
CUDA is mounted has the following steps
(1) CUDA drive itself is mounted
(2) CUDA is mounted cuDNN
(3) configuration settings
Download CUDA
https://developer.nvidia.com/cuda-10.0-download-archive?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal
After the download is complete, double-click operation
GeForce removed
Launched CUDA, if the VS installation, can be installed, if not installed, removed, or easy error
Wait for the installation to complete
After installation, the default path, to see if the installation was successful
If they are present, and the installation is not being given, so that the installation was successful
Here we are then install cuDNN, Download:
https://developer.nvidia.com/rdp/cudnn-download
Once downloaded, unpacked we'll get a folder cuda asking about the price, we will cuda folder renamed, cudnn
Then copy cudnn installation path to go below cuda
Environment variable configuration, this computer, right property,
After installing cuda, will automatically help us to add two environment variables which
We also need to add the path cudnn, and the CUPTI
These few lines specified, must have a 4 lines. Or will be error
We next tested to see if the installation is complete, cmd, enter nvcc -V, the following appears, then the installation is complete, the version number
Today introduced here, next time we continue to learn.
----------------
Disclaimer: This article is CSDN bloggers "Chengdu - God's leg" of the original article, follow the CC 4.0 BY-SA copyright agreement, reproduced, please attach the original source link and this statement.
Original link: https: //blog.csdn.net/u012798683/article/details/96380787