python – modprobe：致命：在目录/ lib / modules /中找不到模块nvidia-uvm

在成功安装和测试使用GPU支持编译的Tensorflow之后,我最近遇到了一个问题.

重新启动计算机后,当我尝试运行Tensorflow程序时收到以下错误消息：

...
('Extracting', 'MNIST_data/t10k-labels-idx1-ubyte.gz')  
modprobe: FATAL: Module nvidia-uvm not found in directory /lib/modules/4.4.0-34-generic
E tensorflow/stream_executor/cuda/cuda_driver.cc:491] failed call to cuInit: CUDA_ERROR_UNKNOWN
I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:140] kernel driver does not appear to be running on this host (caffe-desktop): /proc/driver/nvidia/version does not exist
I tensorflow/core/common_runtime/gpu/gpu_init.cc:92] No GPU devices available on machine.
(0, 114710.45)
(1, 95368.891)
...
(98, 56776.922)
(99, 57289.672)

Screencapture of error

代码：https://github.com/llSourcell/autoencoder_demo

问：为什么重新启动Ubuntu 16.04机器会破坏Tensorflow？

解决方法:

我实际上解决了我自己的问题,并希望分享对我有用的解决方案.

谷歌的神奇搜索是：
“modprobe：致命：在目录/ lib / modules /中找不到模块nvidia-uvm”

这导致我在askubuntu上得到以下答案：
https://askubuntu.com/a/496146

该答案的作者Sneetsher在解释方面做得非常好,如果链接不是404,我会从那里开始.

悬崖笔记

诊断：我怀疑Ubuntu可能在重新启动时安装了内核更新.

解决方案：重新安装NVIDIA驱动程序修复了错误.

问题：无法在运行X服务器的情况下安装NVIDIA驱动程序

修复NVIDIA驱动程序的两种不同方法

1)键盘和显示器：

解释askubuntu答案：

1) Switch to text-only console (Ctrl+Alt+F1 or any to F6).

2) Build driver modules for the current kernel (which just installed) sudo ./<DRIVER>.run -K

信用“Sneetsher”：https://askubuntu.com/a/496146

我没有连接到这台PC的键盘或显示器,所以这是我实际使用的“无头”方法：

2)通过SSH：

按照本指南重新启动到控制台：

http://ubuntuhandbook.org/index.php/2014/01/boot-into-text-console-ubuntu-linux-14-04/

$sudo cp -n /etc/default/grub /etc/default/grub.orig
$sudo nano /etc/default/grub
$sudo update-grub

根据上面的链接编辑grub文件(3个更改)：

1. Comment the line GRUB_CMDLINE_LINUX_DEFAULT=”quiet splash”, by adding # at the beginning, which will disable the Ubuntu purple screen.
2. Change GRUB_CMDLINE_LINUX=”” to GRUB_CMDLINE_LINUX=”text”, this makes Ubuntu boot directly into Text Mode.

3. Uncomment this line #GRUB_TERMINAL=console, by removing the # at the beginning, this makes Grub Menu into real black & white Text Mode (without background image)

4. UPDATE: (If running Ubuntu 16.04 If
   $sudo systemctl set-default multi-user.target

5. Reboot into console

$sudo shutdown -r now
$sudo service lightdm stop
$sudo ./<DRIVER>.run

按照NVIDIA驱动程序安装程序

$sudo mv /etc/default/grub /etc/default/grub.textonly 
$sudo mv /etc/default/grub.orig /etc/default/grub 
$sudo update-grub
$sudo shutdown -r now

结果(现在看起来像GPU成功检测到了什么)

...
('Extracting', 'MNIST_data/t10k-labels-idx1-ubyte.gz')
I tensorflow/core/common_runtime/gpu/gpu_init.cc:118] Found device 0 with properties: 
name: GeForce GTX 970
major: 5 minor: 2 memoryClockRate (GHz) 1.342
pciBusID 0000:01:00.0
Total memory: 3.94GiB
Free memory: 3.88GiB
I tensorflow/core/common_runtime/gpu/gpu_init.cc:138] DMA: 0 
I tensorflow/core/common_runtime/gpu/gpu_init.cc:148] 0:   Y 
I tensorflow/core/common_runtime/gpu/gpu_device.cc:868] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 970, pci bus id: 0000:01:00.0)
(0, 113040.92)
(1, 94895.867)
...

Screencapture of the same