manjaro中tensorflow2支持gpu - Thu, Jun 24, 2021

manjaro中tensorflow2支持gpu

1. 概述

由于系统是manjaro，使用tensroflow进行模型训练，需要支持GPU训练。

2. manjaro安装切换GPU独显

linux使用wine运行游戏

使用以下命令进行独显切换

prime-offload
optimus-manager --switch nvidia

3. manjaro安装cuda,cudnn

sudo pacman -S cuda cuda-tools cudnn

4. 安装tensorflow-gpu

conda install tensorflow-gpu

5. 重启

关键一步

6. 检查

(base) [pana@kaisawind-z2 ~]$ nvidia-smi
Thu Jun 24 11:14:00 2021       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 465.31       Driver Version: 465.31       CUDA Version: 11.3     |
|-------------------------------+----------------------+----------------------+
| 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 |
| N/A   56C    P2    29W /  N/A |    895MiB /  6078MiB |     22%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      3140      G   /usr/lib/Xorg                     304MiB |
|    0   N/A  N/A      3226      G   /usr/bin/kwin_x11                 112MiB |
|    0   N/A  N/A      3288      G   /usr/bin/plasmashell               41MiB |
|    0   N/A  N/A      3976      G   /usr/lib/firefox/firefox          108MiB |
|    0   N/A  N/A      4375      C   ...onda/envs/py38/bin/python      175MiB |
|    0   N/A  N/A      4619      G   ...AAAAAAAAA= --shared-files       35MiB |
|    0   N/A  N/A      6594      G   ...AAAAAAAAA= --shared-files      109MiB |
+-----------------------------------------------------------------------------+

import tensorflow as tf

tf.__version__

print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
print("Num CPUs Available: ", len(tf.config.experimental.list_physical_devices('CPU')))

Output:

Num GPUs Available:  1
Num CPUs Available:  1

device_name = tf.test.gpu_device_name()
print('Found GPU at: {}'.format(device_name))

Output:

Found GPU at: /device:GPU:0