TensorFlow 以及 cuda 和 cudnn 的环境配置问题
https://zhuanlan.zhihu.com/p/351787834
显卡驱动是我们使用显卡的最基础的条件，没有显卡驱动就没有办法使用显卡；但是，对于机器学习或者深度学习环境的配置来说，不必过多的关注显卡驱动，我们要做的只是要看看我们的显卡驱动是不是版本过低，因为每一个版本的 cuda toolkit 都有一个最低版本的显卡驱动要求【如果版本过低或者不是 NVIDIA 驱动，因为CUDA Toolkit本地安装包内含特定版本Nvidia显卡驱动，下面安装 cuda 中可以选择更新显卡驱动】
显卡驱动版本一定不能低，而且支持向后兼容，因此越高越好
如下所示，展示了每一个 cuda 版本对显卡驱动的最低要求，查看：https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html#cuda-major-component-versions
通常来说，我们的显卡版本只要高于这个要求，就可以不在关注显卡驱动的问题
使用 nvidai-smi 命令查看当前机器上的显卡信息：
本机当前的Driver Version: 411.31
所以本机安装的组件的版本：
tensorflow-gpu 1.15.0
cuda_10.0.130_411.31_win10
cudnn-10.0-windows10-x64-v7.6.5.32
1：安装tensorflow 1.15.0, 在激活的环境中安装
如果以上命令不行，也可以使用一下命令尝试
conda install --channel https://conda.anaconda.org/anaconda tensorflow-gpu=1.15.0 或
pip install tensorflow-gpu==1.15.0 -i https://pypi.tuna.tsinghua.edu.cn/simple
2：安装CUDA
CUDA(ComputeUnified Device Architecture)，是显卡厂商NVIDIA推出的运算平台和编程模型。
CUDA是一种由NVIDIA推出的通用并行计算架构，该架构使GPU能够解决复杂的计算问题。允许开发者通过利用图形处理单元（GPU）的功能来进行计算，可以显着提高计算性能。.
如果之前安装了旧的版本，首先可以卸载旧的版本，之前安装的是9.0的版本
首先查看对应关系 ,查看地址：https://tensorflow.google.cn/install/source_windows#gpu
CUDA下载地址：https://developer.nvidia.com/cuda-downloads
下载之前的版本，这里下载的是10.1 update2版本
解压安装，安装完成后查看
C:\Users\Administrator>nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:12:52_Pacific_Daylight_Time_2019
Cuda compilation tools, release 10.1, V10.1.243
3：安装cuDNN
NVIDIA cuDNN是用于深度神经网络的GPU加速库，专门为深度学习网络（如CNN）有了它才能在GPU上完成深度学习的计算它强调性能、易用性和低内存开销。
如果你要用GPU训练模型，cuDNN不是必须的，但是一般会采用这个加速库。
NVIDIA cuDNN可以集成到更高级别的机器学习框架中，如谷歌的Tensorflow、加州大学伯克利分校的流行caffe软件。
简单的插入式设计可以让开发人员专注于设计和实现神经网络模型，而不是简单调整性能，同时还可以在GPU上实现高性能现代并行计算。
cuDnn依赖CUDA，必须要先安装CUDA。
下载地址： https://developer.nvidia.cn/rdp/cudnn-archive 根据版本信息，我们下载7.6.5
下载完成后解压，将cuda中的三个文件夹复制到cuda的安装目录中。
配置pyCharm, 测试代码：
if __name__ == '__main__': os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' tf.compat.v1.disable_eager_execution() print(tf.__version__) print('GPU:', tf.test.is_gpu_available()) a = tf.constant(1.) b = tf.constant(2.) sess = tf.compat.v1.Session() print(sess.run(a + b))
输入结果如下： 红色是程序中print的结果
D:\ProgramData\Anaconda3\envs\tensorflow\python.exe E:/python/anaconda/main.py
2022-01-05 17:20:50.200423: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
1.15.0
2022-01-05 17:20:52.432109: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2022-01-05 17:20:52.435774: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll
2022-01-05 17:20:52.574292: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: GeForce 930M major: 5 minor: 0 memoryClockRate(GHz): 0.941
pciBusID: 0000:01:00.0
2022-01-05 17:20:52.574466: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
2022-01-05 17:20:52.577610: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_100.dll
2022-01-05 17:20:52.580471: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_100.dll
2022-01-05 17:20:52.582051: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_100.dll
2022-01-05 17:20:52.586049: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_100.dll
2022-01-05 17:20:52.588763: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_100.dll
2022-01-05 17:20:52.598200: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2022-01-05 17:20:52.600780: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
GPU: True
2022-01-05 17:20:54.022492: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1159] Device interconnect StreamExecutor with strength 1 edge matrix:
2022-01-05 17:20:54.022637: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1165] 0
2022-01-05 17:20:54.022717: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 0: N
2022-01-05 17:20:54.025966: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/device:GPU:0 with 3048 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:01:00.0, compute capability: 5.0)
2022-01-05 17:20:54.032587: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: GeForce 930M major: 5 minor: 0 memoryClockRate(GHz): 0.941
pciBusID: 0000:01:00.0
2022-01-05 17:20:54.032763: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
2022-01-05 17:20:54.032872: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_100.dll
2022-01-05 17:20:54.032979: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_100.dll
2022-01-05 17:20:54.033083: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_100.dll
2022-01-05 17:20:54.033189: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_100.dll
2022-01-05 17:20:54.033298: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_100.dll
2022-01-05 17:20:54.033405: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2022-01-05 17:20:54.034719: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
2022-01-05 17:20:54.036856: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1618] Found device 0 with properties:
name: GeForce 930M major: 5 minor: 0 memoryClockRate(GHz): 0.941
pciBusID: 0000:01:00.0
2022-01-05 17:20:54.037077: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_100.dll
2022-01-05 17:20:54.037245: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_100.dll
2022-01-05 17:20:54.037369: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_100.dll
2022-01-05 17:20:54.037493: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_100.dll
2022-01-05 17:20:54.037610: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_100.dll
2022-01-05 17:20:54.037730: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_100.dll
2022-01-05 17:20:54.037853: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2022-01-05 17:20:54.045311: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1746] Adding visible gpu devices: 0
2022-01-05 17:20:54.045536: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1159] Device interconnect StreamExecutor with strength 1 edge matrix:
2022-01-05 17:20:54.045677: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1165] 0
2022-01-05 17:20:54.045764: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1178] 0: N
2022-01-05 17:20:54.048339: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1304] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 3048 MB memory) -> physical GPU (device: 0, name: GeForce 930M, pci bus id: 0000:01:00.0, compute capability: 5.0)
3.0
Process finished with exit code 0
出现错误及解决：
1： Attempting to fetch value instead of handling error Internal: failed to get device attribute 13 for device 0: CUDA_ERROR_UNKNOWN: unknown error
问题产生的原因：驱动与cuda不匹配，更新nvidia驱动即可。
解决办法：进入cuda-gpu匹配页面： http://developer.nvidia.com/cuda-gpus
我的显卡是 GeForce 930M
点击，选择930M，
点击 Drivers 链接进入显卡驱动下载页面
=================================
CUDA Device Query ，算力： 5
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\extras\demo_suite>deviceQuery.exe
deviceQuery.exe Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GeForce 930M"
CUDA Driver Version / Runtime Version 10.0 / 10.0
CUDA Capability Major/Minor version number: 5.0
Total amount of global memory: 4096 MBytes (4294967296 bytes)
( 3) Multiprocessors, (128) CUDA Cores/MP: 384 CUDA Cores
GPU Max Clock rate: 941 MHz (0.94 GHz)
Memory Clock rate: 900 Mhz
Memory Bus Width: 64-bit
L2 Cache Size: 1048576 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 4 copy engine(s)
Run time limit on kernels: Yes
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
CUDA Device Driver Mode (TCC or WDDM): WDDM (Windows Display Driver Model)
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: No
Supports Cooperative Kernel Launch: No
Supports MultiDevice Co-op Kernel Launch: No
Device PCI Domain ID / Bus ID / location ID: 0 / 1 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.0, CUDA Runtime Version = 10.0, NumDevs = 1, Device0 = GeForce 930M
Result = PASS
浙公网安备 33010602011771号