【图文教程】TensorFlow-GPU极简-不用安装cuda、cuDNN

前言：

看到网上大部分TensorFlowGPU教程都需要看配置需要装一大堆东西，然后就来更新一篇我知道的最简便的安装GPU版本的方法了，不用安装CUDA CUDNN VS啥的，可能目前内容上有点粗糙，后续会慢慢完善的。个人安装的目的就是为了重新下载一个与keras版本匹配的TensorFlow，具体可以通过这个网址https://docs.floydhub.com/guides/environments/查看Keras与TensorFlow的对应版本，然后我卸载了之前的TensorFlowGPU版本，顺便来记录一下安装过程，当然，读者可以根据自己的需要去下载其他的TensorFlow-GPU

一、准备阶段

1.设备环境：win10+你电脑上有GPU。。。（虽然听起来有点无脑）

2.安装anaconda，这部分就不多说了，需要下载的请直接到官网下载，然后直接安装，不放心的话去看看其他博客教程

3.卸载TensorFlow-GPU+Keras这一步是我自己需要做的，如果你原先没有Tensorflow-gpu就不用管了，直接跳到下一步。

卸载指令：conda uninstall tensorflow-gpu==1.14.0

                  conda uninstall keras==1.0.8

二、安装

1.如果你刚安装好anaconda，或者出于其他原因，第一步就从建立一个虚拟环境开始吧。

按照下图的顺序来，②的命名自己随意，③的选项建议选3.6或7吧，或者有更高版本也好，尽量别选2.X，2.7的据说都快要停止维护了，对python新手而言也不要选版本太高的，可能不稳定。

点击创建之后可能会等一段时间（因为会配备相关必要的库所以可能有点慢）。

2.虚拟环境建完之后，操作如下图：  

再点击“open terminal”进入终端界面：

3.接下来就是重头戏了，安装TensorFlow-GPU：

在终端输入：conda install -c aaronzs tensorflow-gpu

默认应该是下载最新版本，如果要特定版本，就conda install -c aaronzs tensorflow-gpu==xxx

但是我发现这个方法无法下载2.2.0版本，所以我退而求其次地下载了2.2.1了

这个过程可能会报warning什么的，如果可以运行就忽略，有报什么错欢迎在评论区留言。

然后就是漫长地等待了

完成后再输入 conda list看看

4.安装成功！！结束了！！！！对就是这么简单！！！！

但是，最后我们还是测试一下到底有没有配置成功：

因为我用的是Pycharm,这里再附加一下Pycharm配置Anaconda环境的过程：

按照下图所示，或者直接Ctrl+Alt+S

打开设置之后就按照下图操作，选完之后就全点"OK"

操作之后，可以看到右下角会显示当前的环境：

然后关键的一步测试了！操作如下:

新建一个python文件，然后输入以下代码，然后运行：

import tensorflow as tf

import os

#os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

a = tf.constant(1.)

b = tf.constant(2.)

print(a+b)

print(tf.__version__)

print('GPU:', tf.config.list_physical_devices('GPU'))

print(tf.test.is_gpu_available())

输出结果为：

tf.Tensor(3.0, shape=(), dtype=float32)

2.1.0

GPU: [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

True

当然，，由于是用GPU跑的，所以会有这些信息：

2021-01-06 11:54:57.365737: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll

2021-01-06 11:55:00.178440: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll

2021-01-06 11:55:00.862896: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties: 

pciBusID: 0000:01:00.0 name: GeForce GTX 1050 computeCapability: 6.1

coreClock: 1.493GHz coreCount: 5 deviceMemorySize: 2.00GiB deviceMemoryBandwidth: 104.43GiB/s

2021-01-06 11:55:00.863272: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll

2021-01-06 11:55:00.868083: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll

2021-01-06 11:55:00.872729: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_10.dll

2021-01-06 11:55:00.874234: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_10.dll

2021-01-06 11:55:00.879152: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_10.dll

2021-01-06 11:55:00.881920: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_10.dll

2021-01-06 11:55:00.892460: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll

2021-01-06 11:55:00.892777: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0

2021-01-06 11:55:00.893388: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2

2021-01-06 11:55:00.895935: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties: 

pciBusID: 0000:01:00.0 name: GeForce GTX 1050 computeCapability: 6.1

coreClock: 1.493GHz coreCount: 5 deviceMemorySize: 2.00GiB deviceMemoryBandwidth: 104.43GiB/s

2021-01-06 11:55:00.896318: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll

2021-01-06 11:55:00.896510: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll

2021-01-06 11:55:00.896701: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_10.dll

2021-01-06 11:55:00.896891: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_10.dll

2021-01-06 11:55:00.897081: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_10.dll

2021-01-06 11:55:00.897269: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_10.dll

2021-01-06 11:55:00.897458: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll

2021-01-06 11:55:00.897699: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0

2021-01-06 11:55:01.557072: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:

2021-01-06 11:55:01.557299: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102]      0 

2021-01-06 11:55:01.557426: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] 0:   N 

2021-01-06 11:55:01.557723: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1241] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 1335 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1050, pci bus id: 0000:01:00.0, compute capability: 6.1)

然后就结束了！！！完成！！！