How to beginner · 3 min read

How to create a tensor in PyTorch

Q: How to create a tensor in PyTorch

Use torch.tensor() to create a tensor from data, or use factory methods like torch.zeros(), torch.ones(), and torch.randn() to create tensors with specific values or shapes. PyTorch tensors are the core data structure for all computations in torch.

Quick answer

Use torch.tensor() to create a tensor from data, or use factory methods like torch.zeros(), torch.ones(), and torch.randn() to create tensors with specific values or shapes. PyTorch tensors are the core data structure for all computations in torch.

PREREQUISITES

Python 3.8+
pip install torch>=2.0

Setup

Install PyTorch using pip if you haven't already. Use the official command from PyTorch website. For CPU-only:

bash

pip install torch

Step by step

Create tensors from Python lists or arrays using torch.tensor(). Use factory functions like torch.zeros() for zero tensors, torch.ones() for ones, and torch.randn() for random normal tensors.

python

import torch

# Create a tensor from a Python list
tensor_from_list = torch.tensor([[1, 2], [3, 4]])
print("Tensor from list:")
print(tensor_from_list)

# Create a tensor of zeros with shape (2,3)
zeros_tensor = torch.zeros((2, 3))
print("\nZeros tensor:")
print(zeros_tensor)

# Create a tensor of ones with shape (3,2)
ones_tensor = torch.ones(3, 2)
print("\nOnes tensor:")
print(ones_tensor)

# Create a tensor with random values from normal distribution
rand_tensor = torch.randn(2, 2)
print("\nRandom normal tensor:")
print(rand_tensor)

output

Tensor from list:
tensor([[1, 2],
        [3, 4]])

Zeros tensor:
tensor([[0., 0., 0.],
        [0., 0., 0.]])

Ones tensor:
tensor([[1., 1.],
        [1., 1.],
        [1., 1.]])

Random normal tensor:
tensor([[ 0.1234, -0.5678],
        [ 1.2345, -0.9876]])

Common variations

You can create tensors on specific devices like GPU using the device parameter. Also, specify data types with dtype. For example, torch.float32 or torch.int64.

python

import torch

# Create a tensor on GPU if available
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
tensor_gpu = torch.ones((2, 2), device=device)
print(f"Tensor on device: {device}")
print(tensor_gpu)

# Create an integer tensor
int_tensor = torch.tensor([1, 2, 3], dtype=torch.int64)
print("\nInteger tensor:")
print(int_tensor)

output

Tensor on device: cpu
tensor([[1., 1.],
        [1., 1.]])

Integer tensor:
tensor([1, 2, 3])

Troubleshooting

If you get an error like ModuleNotFoundError: No module named 'torch', ensure PyTorch is installed correctly with pip install torch. For device errors, verify your CUDA drivers and PyTorch CUDA version compatibility.

✅

Key Takeaways

Use torch.tensor() to create tensors from data.
Factory methods like torch.zeros() and torch.ones() create tensors with predefined values.
Specify device and data type to control tensor placement and precision.
Always install PyTorch via pip to avoid import errors.
Check CUDA compatibility if using GPU tensors.

Verified 2026-04

Verify ↗