Problem scenario
You want to test PyTorch. What should you do?
Answer
1. Enter a virtaulenv. Use How Do You Install and Create a Virtual Environment? if necessary.
2. Run this: pip3 install torch
3. Use this program:
# This was produced by a Google search in an AI overview result on Aug 4 2025.
# It was slightly modified.
# It has a non-deterministic test of Tensor operations. PyTorch requires a parameter
# to specify CUDA or regular CPU architecture.import torch
def test_pytorch():
"""
Tests the PyTorch installation, including CUDA availability and a basic tensor operation.
"""
print("--- PyTorch Installation Test ---")# 1. Check PyTorch version
print(f"PyTorch version: {torch.__version__}")# 2. Check CUDA availability
if torch.cuda.is_available():
print("CUDA is available! PyTorch can use your GPU.")
print(f"CUDA version: {torch.version.cuda}")
print(f"Number of GPUs: {torch.cuda.device_count()}")
print(f"Current GPU name: {torch.cuda.get_device_name(0)}")
device = torch.device("cuda:0")
else:
print("CUDA is NOT available. PyTorch will run on CPU.")
device = torch.device("cpu")# 3. Perform a basic tensor operation
print(f"\nPerforming a basic tensor operation on {device}...")
try:
x = torch.randn(4, 4, device=device)
y = torch.randn(4, 4, device=device)
z = x + y
print(f"Tensor x:\n{x}")
print(f"Tensor y:\n{y}")
print(f"Tensor z (x + y):\n{z}")
print("Basic tensor operation successful!")
except Exception as e:
print(f"Error during basic tensor operation: {e}")print("\n--- Test Complete ---")
if __name__ == "__main__":
test_pytorch()