Gfilui

I use nvidia card to run deep learning, and no gpu mistakes have occured.
On 2018.02.08, when I run a deep network with caffe-gpu, the mistake occured:(...cudaSuccess...35vs0...).

Then I checked my system setting, found that the Graphic was set to (Skylake GT2), like the picture

However, I ensured I did nothing in GPU setting, and my additional drivers were still " nvidia-384 + microcode ":

After reboot, I run command nvidia-smi,

it printed

command not found

Following some methods from internet, I selected the first graphic choice when run command:
sudo update-alternatives --config x86_64-linux-gnu_gl_conf
and re-run nvidia-smi:

answer has changed to

"NVIDIA-SMI couldn't find libnvidia-ml.so library in your system.
Please make sure that the NVIDIA Display Driver is properly installed
and present in your system. Please also try adding directory that
contains libnvidia-ml.so to your system PATH."

I checked my "nvidia-settings", found there no nvidia or inter alternative:

Finally I found the system load nvidia as 3D controller but not vga compatible controller:

:~$ lspci | grep -i vga

00:02.0 VGA compatible controller: Intel Corporation Skylake Integrated Graphics (rev 06)

~$ lspci | grep -i 3d

01:00.0 3D controller: NVIDIA Corporation GM107M [GeForce GTX 950M] (rev a2)

So, how should I do if I want to use nvidia to run GPU computing?