PyTorch Tensors
PyTorch uses tensors (multi‑dimensional arrays) similar to NumPy, but with GPU acceleration and automatic differentiation. Understanding tensors is the first step to using PyTorch.
Creating Tensors
import torch
# From list
a = torch.tensor([1,2,3])
# Random
b = torch.rand(3,4)
# Zeros
c = torch.zeros(2,3)
# Like NumPy
d = torch.from_numpy(numpy_array)Tensor Operations
Most NumPy operations work on tensors: addition, multiplication, reshaping, indexing.
x = torch.rand(5,3)
print(x.shape)
print(x + x)
print(x.mean(dim=0))Moving to GPU
if torch.cuda.is_available():
device = torch.device('cuda')
x = x.to(device)Autograd (Automatic Differentiation)
PyTorch tracks operations on tensors with `requires_grad=True`. Call `.backward()` to compute gradients.
x = torch.tensor([2.0], requires_grad=True)
y = x**2
y.backward()
print(x.grad) # 2*x = 4.0Two Minute Drill
- Tensors are multi‑dimensional arrays with GPU support.
- Create via `torch.tensor()`, `torch.rand()`, `torch.zeros()`.
- Move to GPU with `.to(device)`.
- Autograd computes gradients automatically.
Need more clarification?
Drop us an email at career@quipoinfotech.com