window11 + python3.7+NVDIA11.7 +pytorchGPU acceleration environment configuration
I have read a lot about pytorch configuring the GPU environment on the Internet. In fact, there is basically no configuration for the Windows 11 version environment, but there is no need to worry about this, and it has no effect.
For bloggers, it’s not like configuring a GPU, because in fact, even if a notebook uses a GPU, the speed is very slow, and it is far from enough for large-scale tasks, that is, doing some small tasks can quickly speed up.
In addition, using a notebook to run deep learning is actually very harmful to the notebook. In many cases, the notebook may be broken. Bloggers really cherish their notebooks, and the main configuration is to learn.
The blogger suggested that try not to use your laptop's GPU to run deep learning. And don't configure it casually, bloggers may be more powerful, they configure it once, and it disappears after using one, you novice configuration configuration may damage the computer environment.
Here is the tutorial:
The blogger has condensed a lot. In fact, the Internet is relatively cumbersome and not that troublesome. There are only a few steps below:
1. Check your own NVDIA version
2. Install the corresponding NVDIA version of cuda and (cudnn can also be installed or not, I see that there is basically no installation on the Internet, this is deep learning optimization acceleration, in fact, cuda can run our calculations on the GPU, and then cudnn It can optimize his calculations, but cudnn is only optimized in terms of deep learning, and cannot qualitatively change the speed, so this can be installed without installation. For example, the three disappearances are now running, and they are running two hours after installation. In fact, this is all Linear reduction of time does not have much impact.) The blogger has not installed cudnn, but if it is a server, try to configure these.
3. Install pytorch corresponding to the cuda version. And then it's over.
Officially begin:
1. Check your own NVDIA version
Enter win+R on the keyboard and enter cmd to open the cmd window:
Enter nvidia-smi to view the NVDIA version
Corresponds to version 11.7.
2. Install the corresponding NVDIA version of cuda
and enter the following URL. Note that the version must be selected.
https://developer.nvidia.com/cuda-11-7-1-download-archive?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_network
After selecting the version, select the configuration
and test whether cuda is installed:
the following is the test, and enter nvcc -V in cmd
3. Install pytorch corresponding to the cuda version.
Enter the following interface:
https://pytorch.org/get-started/locally/
Just install the following Run the command command under cmd, and it's over.
Test whether it is installed:
if the following conditions appear, it is installed.
