ubuntu18.04本地搭建tensorflow-gpu环境

0 前提

1. 须保证ubuntu本机已经安装gcc和nvidia gpu 驱动
2. 须安装好对应版本的cuda和cudnn
3. 安装anaconda3

说明：这些我都之前安装过了，可以参考我之前的博客

1 安装

0 检查条件

root@master:/home/hqc# nvidia-smi
	Tue Dec 21 17:31:19 2021       
	+-----------------------------------------------------------------------------+
	| NVIDIA-SMI 495.44       Driver Version: 495.44       CUDA Version: 11.5     |
	|-------------------------------+----------------------+----------------------+
	| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
	| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
	|                               |                      |               MIG M. |
	|===============================+======================+======================|
	|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
	| 23%   24C    P8     8W / 250W |     11MiB / 11178MiB |      0%      Default |
	|                               |                      |                  N/A |
	+-------------------------------+----------------------+----------------------+
	                                                                               
	+-----------------------------------------------------------------------------+
	| Processes:                                                                  |
	|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
	|        ID   ID                                                   Usage      |
	|=============================================================================|
	|    0   N/A  N/A      1414      G   /usr/lib/xorg/Xorg                  4MiB |
	|    0   N/A  N/A      3477      G   /usr/lib/xorg/Xorg                  4MiB |
	+-----------------------------------------------------------------------------+
# 说明nvidia驱动安装成功

root@master:/home/hqc# python
	Python 3.6.9 (default, Jan 26 2021, 15:33:00) 
	[GCC 8.4.0] on linux
	Type "help", "copyright", "credits" or "license" for more information.
	>>> exit()
# 可查看到gcc版本为8.4.0

1 conda创建虚拟环境

# 查看已创建的虚拟环境
root@master:/home/hqc# conda env list
	conda：未找到命令
# 没有conda命令

解决：

root@master:/home/hqc# cd anaconda3/bin/
root@master:/home/hqc/anaconda3/bin# pwd
	/home/hqc/anaconda3/bin
# 查找anaconda3的路径

# 配置环境变量添加到~/.bashrc文件的最后一行
root@master:/home/hqc# vim ~/.bashrc
	export PATH="/home/hqc/anaconda3/bin:$PATH"
# 使生效
root@master:/home/hqc# source ~/.bashrc
# 查看版本
root@master:/home/hqc# conda --version
	conda 4.5.11

正式创建：

root@master:/home/hqc# conda create -n tf
	Solving environment: done
	
	==> WARNING: A newer version of conda exists. <==
	  current version: 4.5.11
	  latest version: 4.11.0
	
	Please update conda by running
	
	    $ conda update -n base -c defaults conda
	
	## Package Plan ##
	
	  environment location: /home/hqc/anaconda3/envs/tf
	
	Proceed ([y]/n)? y
	
	Preparing transaction: done
	Verifying transaction: done
	Executing transaction: done
	#
	# To activate this environment, use:
	# > source activate tf
	# 激活虚拟环境
	# To deactivate an active environment, use:
	# > source deactivate
	# 退出虚拟环境

# 查看已创建的虚拟环境
root@master:/home/hqc# conda env list
# conda environments:
#
base                  *  /home/hqc/anaconda3
tf                       /home/hqc/anaconda3/envs/tf
# 创建成功

2 虚拟环境中安装tensorflow-gpu

1 安装

# 进入tf虚拟环境
root@master:/home/hqc# source activate tf

# 开始安装
(tf) root@master:/home/hqc# conda install tensorflow-gpu

	Solving environment: done
	
	==> WARNING: A newer version of conda exists. <==
	  current version: 4.5.11
	  latest version: 4.11.0
	
	Please update conda by running
	
	    $ conda update -n base -c defaults conda
	
	## Package Plan ##
	
	  environment location: /home/hqc/anaconda3/envs/tf
	
	  added / updated specs: 
	    - tensorflow-gpu
	
	The following packages will be downloaded:
	
	    package                    |            build
	    ---------------------------|-----------------
	    pip-21.2.4                 |   py39h06a4308_0         2.1 MB
	    tzdata-2021e               |       hda174b7_0         123 KB
	    sqlite-3.36.0              |       hc218d9a_0         1.4 MB
	    libstdcxx-ng-9.1.0         |       hdf63c60_0         4.0 MB
	    aiosignal-1.2.0            |     pyhd3eb1b0_0          12 KB
			...
	    ------------------------------------------------------------
	                                           Total:        1.48 GB
	
	The following NEW packages will be INSTALLED:
	
	    _libgcc_mutex:          0.1-main                
	    _tflow_select:          2.1.0-gpu               
	    absl-py:                0.13.0-pyhd3eb1b0_0     
	    aiohttp:                3.8.1-py39h7f8727e_0    
	    aiosignal:              1.2.0-pyhd3eb1b0_0      
	    astor:                  0.8.1-py39h06a4308_0    
	    astunparse:             1.6.3-py_0              
	    async-timeout:          4.0.1-pyhd3eb1b0_0      
	    attrs:                  21.2.0-pyhd3eb1b0_0     
			...
	Proceed ([y]/n)? 

	Downloading and Extracting Packages
	pip-21.2.4           | 2.1 MB    | ########################################################### | 100% 
	tensorflow-base-2.4. | 422.4 MB  | ########################################################### | 100% 
	cudatoolkit-10.1.243 | 513.2 MB  | ########################################################### | 100% 
	cudnn-7.6.5          | 250.6 MB  | ########################################################### | 100% 
	...
	Preparing transaction: done
	Verifying transaction: done
	Executing transaction: done
# 安装完成

2 验证

(tf) root@master:/home/hqc# python
	Python 3.9.7 (default, Sep 16 2021, 13:09:58) 
	[GCC 7.5.0] :: Anaconda, Inc. on linux
	Type "help", "copyright", "credits" or "license" for more information.
	>>> import tensorflow as tf
		2021-12-21 18:31:55.589944: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1
	>>> import tensorflow as tf
# 配置成功！

3 至此大功告成！

还是挺顺利的。