centos8安装NVIDIA显卡驱动，运行机器学习代码
官方安装文档：NVIDIA Driver Installation Quickstart Guide :: NVIDIA Tesla Documentation
1.查看默认显卡驱动
[root@localhost ~]# lsmod |grep nouveau
nouveau 2351104 0
mxm_wmi 16384 1 nouveau
wmi 32768 2 mxm_wmi,nouveau
video 49152 1 nouveau
i2c_algo_bit 16384 1 nouveau
drm_kms_helper 266240 5 drm_vram_helper,bochs_drm,nouveau
drm_ttm_helper 16384 3 drm_vram_helper,bochs_drm,nouveau
ttm 73728 3 drm_vram_helper,drm_ttm_helper,nouveau
drm 585728 7 drm_kms_helper,drm_vram_helper,bochs_drm,drm_ttm_helper,ttm,nouveau2.安装依赖环境
dnf install -y tar bzip2 make automake gcc gcc-c++ pciutils elfutils-libelf-devel libglvnd-devel
dnf install -y epel-release
distribution=rhel8
ARCH=$( /bin/arch )
dnf config-manager --add-repo http://developer.download.nvidia.com/compute/cuda/repos/$distribution/${ARCH}/cuda-$distribution.repo
dnf install -y kernel-devel-$(uname -r) kernel-headers-$(uname -r)
dnf install -y kernel kernel-core kernel-modules3.安装NVIDIA显卡驱动
dnf clean all
dnf -y module install nvidia-driver:latest-dkms
安装之后，需要重启机器，使显卡驱动生效
[root@localhost ~]# lsmod|grep nouveau
[root@localhost ~]# lsmod|grep nvidia
nvidia_drm 69632 0
nvidia_modeset 1183744 1 nvidia_drm
nvidia_uvm 1339392 0
nvidia 55173120 2 nvidia_uvm,nvidia_modeset
drm_kms_helper 266240 5 drm_vram_helper,bochs_drm,nvidia_drm
drm 585728 8 drm_kms_helper,drm_vram_helper,bochs_drm,nvidia,drm_ttm_helper,nvidia_drm,ttm可以看到，默认显卡驱动已经没了，换成了NVIDIA显卡驱动
查看显卡信息
[root@localhost ~]# nvidia-smi
Tue Oct 25 10:41:05 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 520.61.05 Driver Version: 520.61.05 CUDA Version: 11.8 |
|-------------------------------+----------------------+----------------------+
| 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 | 00000000:00:10.0 Off | N/A |
| 20% 40C P0 N/A / 75W | 0MiB / 4096MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+4.安装NVIDIA CUDA工具包(包含CUPTI)，当前最新为11.8版本,大概下载3.8G
dnf install cuda-toolkit-11-8 -y5.安装NVIDIA CUDNN包,当前最新为11.8版本,需下载681M
dnf install libcudnn8 -y6.安装Python环境进行测试
dnf install python39 python39-devel -y创建虚拟环境，安装tensorflow
python3.9 -m venv py39
source py39/bin/activate
pip install tensorflow通过tensorflow查看可使用设备
python -c 'import tensorflow as tf;print(tf.config.experimental.list_physical_devices())'(py39) [root@localhost ~]# python
Python 3.9.13 (main, Jun 24 2022, 15:32:51)
[GCC 8.5.0 20210514 (Red Hat 8.5.0-13)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
2022-10-25 11:39:46.930430: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-10-25 11:39:47.056206: E tensorflow/stream_executor/cuda/cuda_blas.cc:2981] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2022-10-25 11:39:47.602293: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory
2022-10-25 11:39:47.602366: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory
2022-10-25 11:39:47.602381: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
>>> tf.config.experimental.list_physical_devices()
2022-10-25 11:40:00.293825: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:980] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-10-25 11:40:00.307989: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:980] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-10-25 11:40:00.308201: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:980] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'), PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
测试下GPU速度
1.打开一个窗口，输入命令，实时监控GPU信息
watch -n1 'nvidia-smi'
2.新建test.py文件，输入以下内容，然后运行
import tensorflow as tf
import timeit
def cpu_run():
with tf.device('/cpu:0'):
cpu_a = tf.random.normal([10000, 1000])
cpu_b = tf.random.normal([1000, 2000])
c = tf.matmul(cpu_a, cpu_b)
return c
def gpu_run():
with tf.device('/gpu:0'):
gpu_a = tf.random.normal([10000, 1000])
gpu_b = tf.random.normal([1000, 2000])
c = tf.matmul(gpu_a, gpu_b)
return c
cpu_time = timeit.timeit(cpu_run, number=10)
gpu_time = timeit.timeit(gpu_run, number=10)
print("cpu:", cpu_time, " gpu:", gpu_time)python test.py