TensorFlow安装指南（Centos 7&Windows）及DCGAN demo测试
作者：Kindle君
时间：2017年4月12日。
出处：http://blog.csdn.net/yexiaogu1104/article/details/69055802
声明：版权所有，转载请联系作者并注明出处
本文记录可成功安装TensorFlow的步骤，包括Centos 7 和Windows 10。具体请参考TensorFlow。另外，介绍了一下GitHub的使用。最后，跑通了DCGAN的demo。希望和大家一起交流学习！
1、TensorFlow安装步骤及问题
1.1 linux下使用virtualenv进行安装
| 步骤 | 指令 | 结果或问题 |
|---|---|---|
| 1.通过发出以下命令安装pip和virtualenv： | sudo yum install python-py python-devel python-virtualenv | 1.yum不能使用 –>>检查python配置路径是否正确。 2.centos下yum中不叫python-dev，而是python-devel. |
| 2.通过发出以下命令创建一个virtualenv环境： | sudo virtualenv –system-site-packages /storage/hesiying/hesiying/tensorflow-virtualenv/ | 1.注意使用sudo. 2.安装提示： a.New_python_executable_in_/storage/hesiying/hesiying/tensorflow_virtualenv/bin/python; b.Installing Setuptools……done.; c.Installing Pip……done |
| 3.通过发出以下命令之一激活virtualenv环境： | source /storage/hesiying/hesiying/tensorflow-virtualenv/bin/activate | 成功提示：(tensorflow-virtualenv)[hesiying@localhost hesiying]$ |
| 4.发出以下命令之一在活动的virtualenv环境中安装TensorFlow： | sudo pip install –upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.0.1-cp27-none-linux_x86_64.whl |
1.失败原因： If Step 4 failed (typically because you invoked a pip version lower than 8.1) 2.成功提示：Successfully installed appdirs-1.4.3 funcsigs-1.0.2 mock-2.0.0 numpy-1.12.1 packaging-16.8 pbr-2.0.0 protobuf-3.2.0 pyparsing-2.2.0 setuptools-34.3.3 six-1.10.0 tensorflow-gpu-1.0.1 wheel-0.29.0 |
| 5.测试是否成功 | a.activate your container as step 3. b.python c.>>>import tensorflow as tf d. >>>sess = tf.Session() e.>>>hello = tf.constant(‘Hello,TensorFlow!’) f.>>>sess = tf.Session() g.>>>print(sess.run(hello)) |
成功提示：见下面的代码片段 |
安装成功
(tensorflow-virtualenv)[hesiying@localhost hesiying]$ python
Python 2.7.5 (default, Nov 6 2016, 00:28:07)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-11)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcublas.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcufft.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcurand.so.8.0 locally
>>> hello = tf.constant('Hello,TensorFlow!')
>>> sess = tf.Session()
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:910] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
I tensorflow/core/common_runtime/gpu/gpu_device.cc:885] Found device 0 with properties:
name: GeForce GTX TITAN X
major: 5 minor: 2 memoryClockRate (GHz) 1.076
pciBusID 0000:01:00.0
Total memory: 11.92GiB
Free memory: 4.25GiB
I tensorflow/core/common_runtime/gpu/gpu_device.cc:906] DMA: 0
I tensorflow/core/common_runtime/gpu/gpu_device.cc:916] 0: Y
I tensorflow/core/common_runtime/gpu/gpu_device.cc:975] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX TITAN X, pci bus id: 0000:01:00.0)
>>> print(sess.run(hello))
Hello,TensorFlow!1.2 Windows下使用Anaconda(失败)和 Native pip进行安装
| 步骤 | 指令 | 结果或问题 |
|---|---|---|
| 1.下载并安装Anaconda3- 4.1.1 with python-3.5.2(传说TF 仅仅对python3.5的支持比较好) | 1.按照界面提示安装，记得勾选所有选项. | conda version : 4.1.6 conda-env version : 2.5.1 conda-build version : 1.21.3 python version : 3.5.2.final.0 |
| 2.下载并安装最新的CUDA® Toolkit 8.0或通过百度云盘下载，密码：awwn. | 1.安装路径：B:\Program Files\Navida\CUDA 2.安装过程比较的慢，不要着急哈。安装完之后系统变量会自动为你添加上 3.输入 nvcc -V，输出版本信息则表示成功。如果之前安装了7.5版本的，请卸载后重新安装 |
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2016 NVIDIA Corporation Built on Mon_Jan__9_17:32:33_CST_2017 Cuda compilation tools, release 8.0, V8.0.60 |
| 3.下载并安装cuDNN v5.1 for CUDA 8.0或通过百度云盘下载，密码：awwn. | 将B:\Program Files\Navida\cudnn-8.0-windows10-x64-v5.1\cuda添加到系统环境变量PATH | s |
| 4.用Anaconda安装TensorFlow的GPU版本 | 1.在conda独立空间中创建一个名为tensorflow的环境： conda create -n tensorflow python=3.5.2 2.激活创建的环境: activate tensorflow 3.安装tensorflow的CPU版本: pip install –ignore-installed –upgrade https://storage.googleapis.com/tensorflow/windows/gpu/tensorflow_gpu-1.0.1-cp35-cp35m-win_amd64.whl |
1.在路径\Anaconda3\envs下会出现一个tensorflow的文件夹 2.cmd中会显示(tensorflow) B:\Program Files\Anaconda3> 3.遇到问题：is not a supported wheel on this platform ——>> The Anaconda installation is community supported, not officially supported(马勒个吉的，仅是社区支持，不是官方支持，也就是说在Anaconda中对于whl文件格式不官方支持，那你倒是给出社区支持版的Anaconda啊！！！！！).So I’ve decided to try python 3.5.2 from python.org and this worked immediately1. |
| 5.用Native pip安装TensorFlow | 1.下载安装python 3.5.2 from python.org pip3 install –upgrade tensorflow-gpu 2.验证是否安装成功：python：import tensorflow as tf |
successfully opened CUDA library cublas64_80.dll locally successfully opened CUDA library cudnn64_5.dll locally successfully opened CUDA library cufft64_80.dll locally successfully opened CUDA library nvcuda.dll locally successfully opened CUDA library curand64_80.dll locally |
2、GitHub简易使用指南
介绍如何通过GitHub同步自己的代码。更详细的GitHub操作可以参见廖雪峰的博客。
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git init
重新初始化现存的 Git 版本库于 /storage/hesiying/hesiying/GAN/.git/
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git add DCGAN-tensorflow-master
(tensorflow-virtualenv)[hesiying@localhost GAN]$ ls
DCGAN-tensorflow-master
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git status
# 位于分支 master
无文件要提交，干净的工作区
(tensorflow-virtualenv)[hesiying@localhost GAN]$ touch readm.txt
(tensorflow-virtualenv)[hesiying@localhost GAN]$ ls
DCGAN-tensorflow-master readm.txt
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git add readm.txt
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git commit -m "write a readme.file"
[master a25c2ff] write a readme.file
1 file changed, 0 insertions(+), 0 deletions(-)
create mode 100644 readm.txt
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git status
# 位于分支 master
无文件要提交，干净的工作区
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git remote add origin https://github.com/KindleHe/GAN.git
fatal: 远程 origin 已经存在。
(tensorflow-virtualenv)[hesiying@localhost GAN]$ git push -u origin master
Username for 'https://github.com': KindleHe
Password for 'https://[email protected]':
Counting objects: 98, done.
Delta compression using up to 8 threads.
Compressing objects: 100% (97/97), done.
Writing objects: 100% (98/98), 34.86 MiB | 54.00 KiB/s, done.
Total 98 (delta 1), reused 0 (delta 0)
remote: Resolving deltas: 100% (1/1), done.
To https://github.com/KindleHe/GAN.git
* [new branch] master -> master
分支 master 设置为跟踪来自 origin 的远程分支 master。3、DCGAN demo
验证TensorFlow是否安装成功，同时运行了DCGAN的一个demo程序。具体步骤如下，代码请参照carpedm20的DCGAN-tensorflow。
| 步骤 | 指令 | 问题 | 解决办法 |
|---|---|---|---|
| 1.download dataset | $ python download.py –datasets mnist celebA | 1.ImportError: No module named requests 2.ImportError: No module named tqdm 3. error: unrecognized arguments: –datasets |
1.sudo pip install requests 2.sudo pip install tqdm 3.修改download.py中parser.add_argument第一个参数为“–datasets”。应该是可以接受多个参数，只是下载celebA时连接服务器超时，所以无法使用。 4.这里暂时以MNIST数据及进行demo测试 |
| 2.1.train a model(错误示例) | python main.py –dataset mnist –input_height=28 –output_height=28 –c_dim=1 | 1.ImportError: No module named scipy: Attempting to use uninitialized value generator /g_h0_lin/Matrix |
1.原因：最后缺少–is_train，导致在main.py中line 80:if Flags.is_train:等于默认值false，接下来的dcgan.train(FLAGS)就无法训练，进而导致参数无法被初始化 |
| python main.py –dataset mnist –input_height=28 –output_height=28 –c_dim=1 –is_train | [] Reading checkpoints… [] Failed to find a checkpoint[!] Load failed… 2.Loaded runtime CuDNN library:5005 (compatibility version 5000) but source was compiled with 5110 (compatibility version 5100). |
1.报错语句出现在model.py的train函数中，说明已经进入训练过程。 2.CuDNN库兼容性问题：执行以下指令安装最新的CuDNN sudo cp cuda/include/cudnn.h /usr/local/cuda-8.0/include sudo cp cuda/lib64/libcudnn* /usr/local/cuda-8.0/lib64 sudo chmod a+r /usr/local/cuda-8.0/include/cudnn.h /usr/local/cuda-8.0/lib64/libcudnn* |
|
| 2.2.train a model(错误示例) | 解决方案：python main.py –dataset mnist –input_height=28 –output_height=28 –c_dim=1 –is_train | 无 | 无 |
| test a model | python main.py –dataset mnist –input_height=28 –output_height=28 –c_dim=1 | 无 | 结果参见以下代码 |
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcublas.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcufft.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcurand.so.8.0 locally
{'batch_size': 64,
'beta1': 0.5,
'c_dim': 1,
'checkpoint_dir': 'checkpoint',
'dataset': 'mnist',
'epoch': 25,
'input_fname_pattern': '*.jpg',
'input_height': 28,
'input_width': None,
'is_crop': False,
'is_train': False,
'learning_rate': 0.0002,
'output_height': 28,
'output_width': None,
'sample_dir': 'samples',
'train_size': inf,
'visualize': False}
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:910] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
I tensorflow/core/common_runtime/gpu/gpu_device.cc:885] Found device 0 with properties:
name: GeForce GTX TITAN X
major: 5 minor: 2 memoryClockRate (GHz) 1.076
pciBusID 0000:01:00.0
Total memory: 11.92GiB
Free memory: 9.54GiB
I tensorflow/core/common_runtime/gpu/gpu_device.cc:906] DMA: 0
I tensorflow/core/common_runtime/gpu/gpu_device.cc:916] 0: Y
I tensorflow/core/common_runtime/gpu/gpu_device.cc:975] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX TITAN X, pci bus id: 0000:01:00.0)
---------
Variables: name (type shape) [size]
---------
generator/g_h0_lin/Matrix:0 (float32_ref 110x1024) [112640, bytes: 450560]
generator/g_h0_lin/bias:0 (float32_ref 1024) [1024, bytes: 4096]
generator/g_bn0/beta:0 (float32_ref 1024) [1024, bytes: 4096]
generator/g_bn0/gamma:0 (float32_ref 1024) [1024, bytes: 4096]
generator/g_h1_lin/Matrix:0 (float32_ref 1034x6272) [6485248, bytes: 25940992]
generator/g_h1_lin/bias:0 (float32_ref 6272) [6272, bytes: 25088]
generator/g_bn1/beta:0 (float32_ref 6272) [6272, bytes: 25088]
generator/g_bn1/gamma:0 (float32_ref 6272) [6272, bytes: 25088]
generator/g_h2/w:0 (float32_ref 5x5x128x138) [441600, bytes: 1766400]
generator/g_h2/biases:0 (float32_ref 128) [128, bytes: 512]
generator/g_bn2/beta:0 (float32_ref 128) [128, bytes: 512]
generator/g_bn2/gamma:0 (float32_ref 128) [128, bytes: 512]
generator/g_h3/w:0 (float32_ref 5x5x1x138) [3450, bytes: 13800]
generator/g_h3/biases:0 (float32_ref 1) [1, bytes: 4]
discriminator/d_h0_conv/w:0 (float32_ref 5x5x11x11) [3025, bytes: 12100]
discriminator/d_h0_conv/biases:0 (float32_ref 11) [11, bytes: 44]
discriminator/d_h1_conv/w:0 (float32_ref 5x5x21x74) [38850, bytes: 155400]
discriminator/d_h1_conv/biases:0 (float32_ref 74) [74, bytes: 296]
discriminator/d_bn1/beta:0 (float32_ref 74) [74, bytes: 296]
discriminator/d_bn1/gamma:0 (float32_ref 74) [74, bytes: 296]
discriminator/d_h2_lin/Matrix:0 (float32_ref 3636x1024) [3723264, bytes: 14893056]
discriminator/d_h2_lin/bias:0 (float32_ref 1024) [1024, bytes: 4096]
discriminator/d_bn2/beta:0 (float32_ref 1024) [1024, bytes: 4096]
discriminator/d_bn2/gamma:0 (float32_ref 1024) [1024, bytes: 4096]
discriminator/d_h3_lin/Matrix:0 (float32_ref 1034x1) [1034, bytes: 4136]
discriminator/d_h3_lin/bias:0 (float32_ref 1) [1, bytes: 4]
Total size of variables: 10834690
Total bytes of variables: 43338760
[*] Reading checkpoints...
[*] Success to read DCGAN.model-32002
[*] 0
[*] 1
[*] 2
[*] 3
[*] 4
[*] 5
[*] 6
[*] 7
[*] 8
[*] 9
[*] 10
[*] 11
[*] 12
[*] 13
[*] 14
[*] 15
[*] 16
[*] 17
[*] 18
[*] 19
[*] 20
[*] 21
[*] 22
[*] 23
[*] 24
[*] 25
[*] 26
[*] 27
[*] 28
[*] 29
[*] 30
[*] 31
[*] 32
[*] 33
[*] 34
[*] 35
[*] 36
[*] 37
[*] 38
[*] 39
[*] 40
[*] 41
[*] 42
[*] 43
[*] 44
[*] 45
[*] 46
[*] 47
[*] 48
[*] 49
[*] 50
[*] 51
[*] 52
[*] 53
[*] 54
[*] 55
[*] 56
[*] 57
[*] 58
[*] 59
[*] 60
[*] 61
[*] 62
[*] 63
[*] 64
[*] 65
[*] 66
[*] 67
[*] 68
[*] 69
[*] 70
[*] 71
[*] 72
[*] 73
[*] 74
[*] 75
[*] 76
[*] 77
[*] 78
[*] 79
[*] 80
[*] 81
[*] 82
[*] 83
[*] 84
[*] 85
[*] 86
[*] 87
[*] 88
[*] 89
[*] 90
[*] 91
[*] 92
[*] 93
[*] 94
[*] 95
[*] 96
[*] 97
[*] 98
[*] 99- http://stackoverflow.com/questions/42456461/tensorflow-not-a-supported-wheel-on-this-platform ↩