谷歌colab
- 查看OS版本
!cat /etc/issue
Ubuntu 18.04.2 LTS \n \l
- 查看显卡配置
!nvidia-smi
Thu Aug 8 10:10:22 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.67 Driver Version: 410.79 CUDA Version: 10.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla T4 Off | 00000000:00:04.0 Off | 0 |
| N/A 67C P0 30W / 70W | 237MiB / 15079MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
+-----------------------------------------------------------------------------+
- 查看cuda版本
!nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130
- 查看cudnn版本
!cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2
#define CUDNN_MAJOR 7
#define CUDNN_MINOR 6
#define CUDNN_PATCHLEVEL 2
--
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#include "driver_types.h"
- tensorflow的使用
import tensorflow as tf
print(tf.test.is_gpu_available())
print(tf.__version__)
True
1.14.0
百度aistudio
- 查看OS版本
!cat /etc/issue
Ubuntu 16.04.6 LTS \n \l
- 查看显卡
!nvidia-smi
Thu Aug 8 20:26:10 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 396.37 Driver Version: 396.37 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-SXM2... Off | 00000000:00:07.0 Off | 0 |
| N/A 34C P0 41W / 300W | 0MiB / 16160MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
- 查看cuda版本
!nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Tue_Jun_12_23:07:04_CDT_2018
Cuda compilation tools, release 9.2, V9.2.148
说明
cuda9.2这个版本对任何gpu版本的tensorflow都无法适配。。 期待百度那边早日升级cuda版本到10.0
- 查看cudnn版本
!cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2
#define CUDNN_MAJOR 7
#define CUDNN_MINOR 3
#define CUDNN_PATCHLEVEL 1
--
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#include "driver_types.h"
- tensorflow的使用(需手动安装)
!pip install tensorflow-gpu
说明
- 不指定版本的话，默认下载最新版本。import tensorflow不报错，但无可用的gpu，tensorflow检测不到相应版本的cuda包，自动切换到cpu版本
- !pip install tensorflow-gpu==1.12.0 后，import tensorflow 报错，提示需要cuda9.0，实际已装cuda9.2，cuda版本不匹配
- 高版本(1.14.0/1.13.1)的gpu需要cuda10.0，低版本(1.5.0-1.12.0)的gpu需要cuda9.0
import tensorflow as tf
print(tf.test.is_gpu_available())
print(tf.__version__)
False
1.14.0
总结对比
- colab环境较新，对tensorflow支持的较好。但是不支持数据的存储，断开重连后数据会丢失
- aistudio支持数据的存储，断开重连后数据还在。但运行环境只对自家的paddle适配的好，不自带tensorflow的包，每次重启需重新下载，且目前cuda的版本无法适配任何一版gpu的tensorflow