CUDA availability check in PyTorch - ML Experiment: Train & Evaluate

Experiment - CUDA availability check

Problem:You want to check if your computer can use the GPU with CUDA to speed up PyTorch computations.

Current Metrics:N/A - This is a hardware/software environment check, not a model training task.

Issue:You are unsure if CUDA is available and properly configured for PyTorch on your system.

Your Task

Write a simple PyTorch script to check if CUDA is available and print the result clearly.

Use only PyTorch library functions.

Do not attempt to train any model.

Output must be clear and user-friendly.

Hint 1

Hint 2

Hint 3

Solution

PyTorch

import torch

def check_cuda():
    if torch.cuda.is_available():
        num_gpus = torch.cuda.device_count()
        print(f"CUDA is available! Number of GPUs detected: {num_gpus}")
        for i in range(num_gpus):
            print(f"GPU {i}: {torch.cuda.get_device_name(i)}")
    else:
        print("CUDA is not available on this system.")

if __name__ == "__main__":
    check_cuda()

Added a function to check CUDA availability using torch.cuda.is_available().

Printed the number of GPUs detected with torch.cuda.device_count().

Printed the name of each GPU detected with torch.cuda.get_device_name(i).

Added a clear message if CUDA is not available.

Results Interpretation

Before: User did not know if CUDA was available. After: User runs the script and sees a clear message about CUDA availability and GPU details.

Checking CUDA availability is a simple but important first step to ensure your PyTorch code can run on GPU hardware for faster training and inference.

Bonus Experiment

Modify the script to create a tensor on the GPU if CUDA is available, and on CPU otherwise, then print the device of the tensor.

💡 Hint

Use torch.device('cuda') or torch.device('cpu') and create a tensor with .to(device). Then print tensor.device.