CUDA Configuration for Windows
Step 1: NVIDIA Video Driver
You should install the latest version of your GPUs driver. - Check which GPU is present in your system
You can download drivers here:
How to check if already present
1. Device Manager Press Win + X
2. click Device Manager
3. Expand Display adapters
4. Look for something like: NVIDIA GeForce / RTX / GTX…
Step 2: Visual Studio C++
You will need Visual Studio, with C++ installed.
🛑IMP: By default, C++ is not installed with Visual Studio, so make sure you select ALL of the C++ options.
Step 3: Anaconda/Miniconda
You will need anaconda to install all deep learning packages
Step 4: CUDA Toolkit
Which version to choice:
Go to https://pytorch.org/get-started/locally/
Check which stable version is available for pytorch: example CUDA 11.8 and CUDA 12.1 but not CUDA 12.4
Recommended as per following image: So you can ether install 11.8 or 12.1 [If you plan to use pytorch]
How to verify if downloaded or not: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA Here you will see the downloaded version folder
Step 5: cuDNN
GPU accelerated library of primitives for deep neural networks
Download the version which is suitable for your CUDA version
Next Step:
Unzip the downlaoded folder
Copy all bin folder files from cuDNN to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8 bin folder --- [Note I have CUDA 12.8 so by dir path include v12.8 check yours]
same copy paste all files for include and lib folder too
How to verify if everything is correctly done
Check for the environment variable set or not
If it is set automatically then everything is perfect if not then set manually
Step 6 Verification if system is using CUDA
Install PyTorch
select the cuda version you installed and all other options as per requirement
Recommendation: You can test this in virtual env too
Run the following script to test your GPU
import torch
print("Number of GPU: ", torch.cuda.device_count())
print("GPU Name: ", torch.cuda.get_device_name())
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)
-x- 😊 THANK YOU 😊 -x-
