What is a PyTorch tensor and how does it differ from a NumPy array?

A tensor is PyTorch's core data structure — an n-dimensional array similar to NumPy's ndarray, but with two critical extra capabilities: it can live on a GPU for accelerated computation, and it supports automatic differentiation (autograd) for computing gradients during backpropagation.

import torch
import numpy as np

# Creating tensors
t1 = torch.tensor([1.0, 2.0, 3.0])          # from Python list
t2 = torch.zeros(3, 4)                       # 3×4 zeros
t3 = torch.ones(2, 3)                        # 2×3 ones
t4 = torch.rand(2, 3)                        # uniform random [0,1)
t5 = torch.randn(2, 3)                       # standard normal
t6 = torch.arange(0, 10, 2)                  # [0, 2, 4, 6, 8]
t7 = torch.linspace(0, 1, 5)                 # 5 evenly spaced pts

# Shape, dtype, device
print(t2.shape)     # torch.Size([3, 4])
print(t1.dtype)     # torch.float32
print(t1.device)    # cpu

# NumPy ↔ PyTorch bridge (shares memory on CPU!)
np_array = np.array([1.0, 2.0, 3.0])
torch_from_np = torch.from_numpy(np_array)   # shares memory
np_from_torch = t1.numpy()                   # shares memory

np_array[0] = 99
print(torch_from_np[0])  # tensor(99.) — memory is shared!