赞
踩
满心欢喜的以为用anaconda3 5.3.1装好了tensorflow-gpu
在vscode开用的时候,一盆冷水扑过来
PS D:\AI_program\yolo3-keras-master> E:\anaconda3\envs\tensorflow-gpu\python.exe
Python 3.7.0 (default, Jun 28 2018, 08:04:48) [MSC v.1912 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
2020-02-27 14:47:07.038599: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
【本机环境】
Python 3.7.0
tensorflow-gpu 2.1.0
keras 2.3.1
cudnn-10.0-windows10-x64-v7.6.4.38
cuda_10.0.130_411.31_win10
看到网上说tensorflow2.0对应的cuda版本是10.0,绝对不能下载错,那我就试试吧
1、安装CUDA
下载地址:https://developer.nvidia.com/cuda-10.0-download-archive?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal
安装的时候关联了VScode
2、cuDNN的安装
先去官网注册,注册真慢到想哭,还要进行recaptcha测试,不经意间进行了一次图灵测试,和人工智能鼻祖进了一步
下载地址https://developer.nvidia.com/rdp/cudnn-archive
帮大家避坑,直接分享网盘
链接: https://pan.baidu.com/s/16icf592tLGFlXN8AhHzT2A 提取码: qgvs
3、环境变量设置
到环境变量中的系统变量的path下默认有第1和第2个增加第3和第4个
C:\Users\shan-dell>nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:04_Central_Daylight_Time_2018
Cuda compilation tools, release 10.0, V10.0.130
4、仍然存在问题
2020-02-27 16:25:15.357944: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library ‘cudart64_101.dll’; dlerror: cudart64_101.dll not found
按网上的说说法备份一个cudart64_100.dll,改一个cudart64_101.dll
顺利完成
(tensorflow-gpu) E:\>python
Python 3.7.0 (default, Jun 28 2018, 08:04:48) [MSC v.1912 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow
2020-02-27 16:38:16.817624: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。