python系列&deep_study系列：解决nvidia-smi “Failed to initialize NVML: Driver/library version mismatch”

本文链接：https://blog.youkuaiyun.com/weixin_54626591/article/details/142971117

解决nvidia-smi “Failed to initialize NVML: Driver/library version mismatch”

nvidia-smi 错误信息

nvidia-smi NVIDIA System Management Interface是用来查看NVIDIA GPU设备的一个命令行工具。今天发现有如下错误。

# 错误信息
myid@[hostname]:~$ nvidia-smi
Failed to initialize NVML: Driver/library version mismatch
NVML library version: 535.146

分析过程

这个问题还是挺常见的，在此主要记录一下我遇到的具体情况。

查看升级日志

在2024-01-11 发生了升级操作。不是工作时段，很可能是自动完成的。

myid@[hostname]:~$ cat /var/log/dpkg.log |grep nvidia|grep libnvidia-common
2024-01-11 06:55:01 upgrade libnvidia-common-535:all 535.129.03-0ubuntu0.22.04.1 535.146.02-0ubuntu0.22.04.1
2024-01-11 06:55:01 status half-configured libnvidia-common-535:all 535.129.03-0ubuntu0.22.04.1
2024-01-11 06:55:01 status unpacked libnvidia-common-535:all 535.129.03-0ubuntu0.22.04.1
2024-01-11 06:55:01 status half-installed libnvidia-common-535:all 535.129.03-0ubuntu0.22.04.1
2024-01-11 06:55:01 status unpacked libnvidia-common-535:all 535.146.02-0ubuntu0.22.04.1
2024-01-11 06:55:01 configure libnvidia-common-535:all 535.146.02-0ubuntu0.22.04.1 <none>
2024-01-11 06:55:01 status unpacked libnvidia-common-535:all 535.146.02-0ubuntu0.22.04.1
2024-01-11 06:55:01 status half-configured libnvidia-common-535:all 535.146.02-0ubuntu0.22.04.1
2024-01-11 06:55:01 status installed libnvidia-common-535:all 535.146.02-0ubuntu0.22.04.1

查看当前运行的driver version

当前运行的driver version还是535.129.03。

# driver version还是535.129
myid@[hostname]:~$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module  535.129.03  Thu Oct 19 18:56:32 UTC 2023
GCC version:

查看已安装的driver version

可以看到其实driver 535.146.02已经安装了，只是运行的还是旧版本。

# driver version 535.146已经安装
myid@[hostname]:~$ apt list --installed |grep -i nvidia-driver

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

nvidia-driver-535/now 535.146.02-0ubuntu0.22.04.1 amd64 [installed,local]

根据网上类似问题的线索，大概率是自动升级后library立即更新，而driver由于是kernal module不会自动运行新版本导致两者不匹配。

所以后续最好关闭自动更新。

解决

reboot

什么也不需要做，reboot 后 driver version 就更新了。在下文可以看到reboot后driver已更新到535.146.02 ，和NVML保持一致了。

myid@[hostname]:~$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module  535.146.02  Sun Dec  3 14:06:14 UTC 2023
GCC version:  gcc version 12.3.0 (Ubuntu 12.3.0-1ubuntu1~22.04)

nvidia-smi也可以正常使用了。

myid@[hostname]:~$ nvidia-smi
Fri Jan 26 14:08:38 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.146.02             Driver Version: 535.146.02   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| 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 RTX 4080        Off | 00000000:65:00.0 Off |                  N/A |
| 57%   28C    P8              13W / 320W |      9MiB / 16376MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|    0   N/A  N/A      1377      G   /usr/lib/xorg/Xorg                            4MiB |
+---------------------------------------------------------------------------------------+