windows下搭建tensorflow开发环境

本文详细介绍了如何在Windows环境下为TensorFlow 1.15配置GPU支持,包括CUDA 10.0和cuDNN v7.6.5的安装步骤,Anaconda环境的创建及TensorFlow的正确安装方法。

摘要生成于 C知道 ,由 DeepSeek-R1 满血版支持, 前往体验 >

1 安装cuda 10.0

下载地址:https://developer.nvidia.com/cuda-toolkit-archive

参考文档:https://blog.youkuaiyun.com/weixin_44307764/article/details/94909104

https://blog.youkuaiyun.com/weixin_45023983/article/details/99178625

cuda版本跟tensorflow版本要对应。 tensorflow1.15 对应cud10

 

2 cuDNN下载 Download cuDNN v7.6.5 (November 18th, 2019), for CUDA 10.2

下载深度神经网络加速库。版本必须与cuda对应。

https://developer.nvidia.com/rdp/cudnn-download

下载后将文件目录,拷贝到cuda安装目录:

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2\      +  cuDNN\bin..

3 安装anaconda。

https://repo.anaconda.com/archive/Anaconda3-2019.10-Windows-x86_64.exe

安装简单, 完成后cmd, 输入python  或ipython测试。

4 Anaconda下创建tensorflow环境。

参考:https://blog.youkuaiyun.com/XunCiy/article/details/89069193

Anaconda默认环境root环境下使用的是python3.7,  目前 tensorflow最新支持的python为3.6,

创建一个tensorflow的环境, 使用python版本是3.6.5

conda create --name tensorflow python=3.6.5 

安装完成后可以看到base环境下多了一个环境tensorflow

 

5 安装tensorflow

打开anaconda prompt命令行,请确保输入activate tensorflow后进入了tensorflow环境

安装指定GPU版本:pip install --ignore-installed --upgrade tensorflow-gpu==1.15,必须制定版本,否则安装最新版本,容易出现兼容性问题

下属对应图,实际测试,不是很准确,基本在附近。还是要实测。

 

PathCompilerCUDA/cuDNNSIMDNotes
2.0.0\py37\CPU\sse2VS2019 16.3Nox86_64Python 3.7
2.0.0\py37\CPU\avx2VS2019 16.3NoAVX2Python 3.7
2.0.0\py37\GPU\cuda101cudnn76sse2VS2019 16.310.1.243_426.00/7.6.4.38x86_64Python 3.7/Compute 3.0
2.0.0\py37\GPU\cuda101cudnn76avx2VS2019 16.310.1.243_426.00/7.6.4.38AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.15.0\py37\CPU+GPU\cuda101cudnn76sse2VS2019 16.310.1.243_426.00/7.6.4.38x86_64Python 3.7/Compute 3.0
1.15.0\py37\CPU+GPU\cuda101cudnn76avx2VS2019 16.310.1.243_426.00/7.6.4.38AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.14.0\py37\CPU\sse2VS2019 16.1Nox86_64Python 3.7
1.14.0\py37\CPU\avx2VS2019 16.1NoAVX2Python 3.7
1.14.0\py37\GPU\cuda101cudnn76sse2VS2019 16.110.1.168_425.25/7.6.0.64x86_64Python 3.7/Compute 3.0
1.14.0\py37\GPU\cuda101cudnn76avx2VS2019 16.110.1.168_425.25/7.6.0.64AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.13.1\py37\CPU\sse2VS2017 15.9Nox86_64Python 3.7
1.13.1\py37\CPU\avx2VS2017 15.9NoAVX2Python 3.7
1.13.1\py37\GPU\cuda101cudnn75sse2VS2017 15.910.1.105_418.96/7.5.0.56x86_64Python 3.7/Compute 3.0
1.13.1\py37\GPU\cuda101cudnn75avx2VS2017 15.910.1.105_418.96/7.5.0.56AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.12.0\py36\CPU\sse2VS2017 15.8Nox86_64Python 3.6
1.12.0\py36\CPU\avx2VS2017 15.8NoAVX2Python 3.6
1.12.0\py36\GPU\cuda100cudnn73sse2VS2017 15.810.0.130_411.31/7.3.1.20x86_64Python 3.6/Compute 3.0
1.12.0\py36\GPU\cuda100cudnn73avx2VS2017 15.810.0.130_411.31/7.3.1.20AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.12.0\py37\CPU\sse2VS2017 15.8Nox86_64Python 3.7
1.12.0\py37\CPU\avx2VS2017 15.8NoAVX2Python 3.7
1.12.0\py37\GPU\cuda100cudnn73sse2VS2017 15.810.0.130_411.31/7.3.1.20x86_64Python 3.7/Compute 3.0
1.12.0\py37\GPU\cuda100cudnn73avx2VS2017 15.810.0.130_411.31/7.3.1.20AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.11.0\py36\CPU\sse2VS2017 15.8Nox86_64Python 3.6
1.11.0\py36\CPU\avx2VS2017 15.8NoAVX2Python 3.6
1.11.0\py36\GPU\cuda100cudnn73sse2VS2017 15.810.0.130_411.31/7.3.0.29x86_64Python 3.6/Compute 3.0
1.11.0\py36\GPU\cuda100cudnn73avx2VS2017 15.810.0.130_411.31/7.3.0.29AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.11.0\py37\CPU\sse2VS2017 15.8Nox86_64Python 3.7
1.11.0\py37\CPU\avx2VS2017 15.8NoAVX2Python 3.7
1.11.0\py37\GPU\cuda100cudnn73sse2VS2017 15.810.0.130_411.31/7.3.0.29x86_64Python 3.7/Compute 3.0
1.11.0\py37\GPU\cuda100cudnn73avx2VS2017 15.810.0.130_411.31/7.3.0.29AVX2Python 3.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0,7.5
1.10.0\py36\CPU\sse2VS2017 15.8Nox86_64Python 3.6
1.10.0\py36\CPU\avx2VS2017 15.8NoAVX2Python 3.6
1.10.0\py36\GPU\cuda92cudnn72sse2VS2017 15.89.2.148.1/7.2.1.38x86_64Python 3.6/Compute 3.0
1.10.0\py36\GPU\cuda92cudnn72avx2VS2017 15.89.2.148.1/7.2.1.38AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.10.0\py27\CPU\sse2VS2017 15.8Nox86_64Python 2.7
1.10.0\py27\CPU\avx2VS2017 15.8NoAVX2Python 2.7
1.10.0\py27\GPU\cuda92cudnn72sse2VS2017 15.89.2.148.1/7.2.1.38x86_64Python 2.7/Compute 3.0
1.10.0\py27\GPU\cuda92cudnn72avx2VS2017 15.89.2.148.1/7.2.1.38AVX2Python 2.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.9.0\py36\CPU\sse2VS2017 15.7Nox86_64Python 3.6
1.9.0\py36\CPU\avx2VS2017 15.7NoAVX2Python 3.6
1.9.0\py36\GPU\cuda92cudnn71sse2VS2017 15.79.2.148/7.1.4x86_64Python 3.6/Compute 3.0
1.9.0\py36\GPU\cuda92cudnn71avx2VS2017 15.79.2.148/7.1.4AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.9.0\py27\CPU\sse2VS2017 15.7Nox86_64Python 2.7
1.9.0\py27\CPU\avx2VS2017 15.7NoAVX2Python 2.7
1.9.0\py27\GPU\cuda92cudnn71sse2VS2017 15.79.2.148/7.1.4x86_64Python 2.7/Compute 3.0
1.9.0\py27\GPU\cuda92cudnn71avx2VS2017 15.79.2.148/7.1.4AVX2Python 2.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.8.0\py36\CPU\sse2VS2017 15.4Nox86_64Python 3.6
1.8.0\py36\CPU\avx2VS2017 15.4NoAVX2Python 3.6
1.8.0\py36\GPU\cuda91cudnn71sse2VS2017 15.49.1.85.3/7.1.3x86_64Python 3.6/Compute 3.0
1.8.0\py36\GPU\cuda91cudnn71avx2VS2017 15.49.1.85.3/7.1.3AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.8.0\py27\CPU\sse2VS2017 15.4Nox86_64Python 2.7
1.8.0\py27\CPU\avx2VS2017 15.4NoAVX2Python 2.7
1.8.0\py27\GPU\cuda91cudnn71sse2VS2017 15.49.1.85.3/7.1.3x86_64Python 2.7/Compute 3.0
1.8.0\py27\GPU\cuda91cudnn71avx2VS2017 15.49.1.85.3/7.1.3AVX2Python 2.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.7.0\py36\CPU\sse2VS2017 15.4Nox86_64Python 3.6
1.7.0\py36\CPU\avx2VS2017 15.4NoAVX2Python 3.6
1.7.0\py36\GPU\cuda91cudnn71sse2VS2017 15.49.1.85.3/7.1.2x86_64Python 3.6/Compute 3.0
1.7.0\py36\GPU\cuda91cudnn71avx2VS2017 15.49.1.85.3/7.1.2AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.7.0\py27\CPU\sse2VS2017 15.4Nox86_64Python 2.7
1.7.0\py27\CPU\avx2VS2017 15.4NoAVX2Python 2.7
1.7.0\py27\GPU\cuda91cudnn71sse2VS2017 15.49.1.85.3/7.1.2x86_64Python 2.7/Compute 3.0
1.7.0\py27\GPU\cuda91cudnn71avx2VS2017 15.49.1.85.3/7.1.2AVX2Python 2.7/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.6.0\py36\CPU\sse2VS2017 15.4Nox86_64Python 3.6
1.6.0\py36\CPU\avx2VS2017 15.4NoAVX2Python 3.6
1.6.0\py36\GPU\cuda91cudnn71sse2VS2017 15.49.1.85.3/7.1.1x86_64Python 3.6/Compute 3.0
1.6.0\py36\GPU\cuda91cudnn71avx2VS2017 15.49.1.85.3/7.1.1AVX2Python 3.6/Compute 3.0,3.5,5.0,5.2,6.1,7.0
1.6.0\py27\CPU\sse2VS2017 15.4Nox86_64Python 2.7

 

 

 

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值