Caffe is already running because it has been installed Cuda7.0+caffe on Ubuntu 14.04 x64 and has been configured well.
However, with the recent need to use torch, and the torch cuda requirement is 8.0, it is decided to upgrade the Cuda version to meet the needs of the torch platform.
The latest version of Caffe has also supported cuda8.0.
Don't say much, start installing cuda8.0.
Graphics: GeForce GTX TITAN x System: Ubuntu 14.04 (x_64) cuda:cuda_8.0.61_375.26_linux.runcudnn:cudnn-8.0-linux-x64-v5.1.tgz
1. The GeForce GTX TITAN x graphics card must be in Cuda version 8.0. Cuda download from here. Remember, do not download the Deb package, or the rear of countless pits waiting for you.
In the installation
GTX1080 graphics card must be cudnn-8.0-v5.1 version, otherwise with Caffe run model, with CPU or GPU graphics running accuracy normal, once opened CUDNN mode, Precision (ACC) immediately down to about 0.1, loss very large. Download the CUDNN here. Download is required to register. It is best to register an account, select the corresponding version, do not use the online other tutorials to the ready-made package, the probability of a problem is very large.
Note: Select No when you are prompted to install the Nvidia driver during the installation. Other options yes or default.
Do your accept the previously read Eula?accept/decline/quit:acceptinstall NVIDIA accelerated Graphics Driver for linux-x86 _64 375.26? (y) es/(n) o/(q) uit:ninstall the CUDA 8.0 Toolkit? (y) es/(n) o/(q) uit:yenter Toolkit location[default is/usr/local/cuda-8.0]:D o want to install a symbolic link at/us R/local/cuda? (y) es/(n) o/(q) uit:yinstall the CUDA 8.0 Samples? (y) es/(n) o/(q) uit:yenter cuda Samples location[default Is/home/zhou]:installing the CUDA Toolkit in/usr/local/cuda-8. 0 ...
You can see the following message when you are finished:
Driver:not selectedtoolkit:installed in/usr/local/cuda-8.0samples:installed In/home/startag, but missing recommended Libraries
If the cuda8.0 video driver is installed during the installation process, this will conflict with the original Nvidia graphics driver that was installed during the restart, and there will always be an icon flashing when booting.
At this time CTRL + ALT +f1~7 are not responding, this is because the two version of the graphics driver does not match, the conflict caused.
Solutions are:
Restart (reboot), ESC---Ubuntu Recovery mode
After waiting for a while, enter a selection screen, select root mode, and after entering command line mode, use the following command to uninstall the NVIDIA graphics driver:
sudo apt-get purge nvidia-*
The Nvidia is all unloaded, and then reboot can go in.
If Apt-get is not available during the use of the above command, you will need to fix the broken package with fix mode, and then uninstall the Nvidia graphics driver after the repair is complete before entering root mode.
Solution See:Ubuntu GNOME 16.04, after installing NVIDIA driver can not boot, how to solve?
The CUDA program can be run normally after a reboot.
Cuda 8.0 and Sample installation check:
After the installation is complete, you need to check that the CUDA program is installed successfully:
Enter/usr/local/cuda/samples and execute the following command to build samples
sudo make all-j8
Once the compilation is complete, enter./bin/x86_64/linux/release, run Devicequery
./devicequery
If the following information appears,
static= FAIL
You need to use the chmod 777-r Devicequery method to give the Devicequery file sudo permissions, the solution see:CUDA devicequery returned error after upgrade to n Vidia 334.21-1
The following information appears, indicating that Cuda and Cuda sample installation succeeded:
Cuda device Query (Runtime API) version (Cudart static linking) detected 1 CUDA capable device (s) device 0: "GeForce GTX Tit An X "Cuda Driver version/runtime version 8.0/8.0 cuda Capability major/minor version number:5.2 total Amount of global memory:12199 MBytes (12791185408 bytes) (multiprocessors) CUDA CORES/MP: 3072 CUDA cores GPU Max clock rate:1076 MHz (1.08 GHz) Memory Clock rate: 3505 Mhz Memory Bus width:384-bit L2 Cache Size: 3145728 bytes Maximum Texture Dimension Size (x, Y, z) 1d= (65536), 2d= (65536, 65536), 3d= (4096, 4096, 4096) Maxi Mum layered 1D Texture size, (num) layers 1d= (16384), 2048 layers Maximum layered 2D Texture Size, (num) layers 2d= (163 (16384), 2048 layers total amount of constant memory:65536 bytes Total amount of shared memory per Blo ck:49152 bytes TOtal number of registers available per block:65536 Warp size:32 Maximum number of Threads per multiprocessor:2048 Maximum number of threads per block:1024 Max dimension size of a thread blo CK (x, Y, z): Max dimension size of a grid size (x, Y, z): (2147483647, 65535, 65535) Maximum memory Pit ch:2147483647 bytes Texture alignment:512 bytes Concurrent Copy a nd kernel execution:yes with 2 copy engine (s) Run time limit on Kernels:yes Integrated GP U Sharing host Memory:no support host page-locked Memory Mapping:yes Alignment requirement for SURFAC Es:yes device has ECC support:disabled device supports Unified Addressing (UVA): Yes Device PCI Domain id/bus id/location id:0/1/0 Compute Mode: < Default (multiple host threads can Use:: Cudasetdevice () WITH device simultaneously) >devicequery, cuda Driver = Cudart, cuda Driver version = 8.0, cuda Runtime version = 8.0, N Umdevs = 1, Device0 = GeForce GTX TITAN xresult = PASS
You can also use the following command to test whether the installation was successful:
Nvidia-smi
Output information:
Thu 20:33:01 +-----------------------------------------------------------------------------+| Nvidia-smi 375.26 Driver version:375.26 | | -------------------------------+----------------------+----------------------+| GPU Name persistence-m| Bus-id Disp.a | Volatile Uncorr. ECC | | Fan Temp Perf pwr:usage/cap| Memory-usage | Gpu-util Compute M. | | ===============================+======================+======================|| 0 GeForce GTX Tit ... Off | 0000:01:00.0 on | N/A | | 22% 40C P8 17w/250w | 500mib/12198mib | 0% Default |+-------------------------------+----------------------+----------------------+ +-------------------------------------------------------------------- ---------+| Processes:gpu Memory | | GPU PID TypE Process name Usage | | =============================================================================|| 0 1275 g/usr/lib/xorg/xorg 267MiB | | 0 2416 G Compiz 106MiB | | 0 2978 G ... el-token=efac54a3cb4fc0dbaf418394276e4c3b 124MiB |+----------------------------------------------- ------------------------------+
Ubuntu 14.04 (x64) installation cuda8.0