What is PyTorch?

It’s a Python-based scientific computing package targeted at two sets of audiences:

• A replacement for NumPy to use the power of GPUs
• a deep learning research platform that provides maximum flexibility and speed

Getting Started

Tensors

Tensors are similar to NumPy’s ndarrays, with the addition being that Tensors can also be used on a GPU to accelerate computing.

from __future__ import print_function
import torch

Construct a 5x3 matrix, uninitialized:

x = torch.empty(5, 3)
print(x)

Out:

tensor([[1.4745e-35, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 1.4013e-45, 0.0000e+00],
        [1.4013e-45, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00]])

Construct a randomly initialized matrix:

x = torch.rand(5, 3)
print(x)

Out:

tensor([[0.8144, 0.2020, 0.0938],
        [0.5893, 0.7267, 0.0332],
        [0.9758, 0.8942, 0.9941],
        [0.7851, 0.9456, 0.1738],
        [0.7761, 0.7344, 0.8812]])

Construct a matrix filled zeros and of dtype long:

x = torch.zeros(5, 3, dtype=torch.long)
print(x)

Out:

tensor([[0, 0, 0],
        [0, 0, 0],
        [0, 0, 0],
        [0, 0, 0],
        [0, 0, 0]])

Construct a tensor directly from data:

x = torch.tensor([5.5, 3])
print(x)

Out:

tensor([5.5000, 3.0000])

or create a tensor based on an existing tensor. These methods will reuse properties of the input tensor, e.g. dtype, unless new values are provided by user

x = x.new_ones(5, 3, dtype=torch.double)      # new_* methods take in sizes
print(x)

x = torch.randn_like(x, dtype=torch.float)    # override dtype!
print(x)                                      # result has the same size

Out:

tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]], dtype=torch.float64)
tensor([[-0.3969, -0.2255, -0.2663],
        [ 0.0611, -1.5954, -1.0884],
        [ 1.3799,  2.5789, -1.0771],
        [ 0.2965,  2.2555, -0.2628],
        [-1.3045,  1.6418,  0.8463]])

Get its size:

print(x.size())

Out:

torch.Size([5, 3])

Note

torch.Size is in fact a tuple, so it supports all tuple operations.

Operations

There are multiple syntaxes for operations. In the following example, we will take a look at the addition operation.

Addition: syntax 1

y = torch.rand(5, 3)
print(x + y)

Out:

tensor([[-0.2879,  0.6880,  0.2436],
        [ 0.4299, -0.7803, -1.0528],
        [ 1.5577,  2.9404, -0.5684],
        [ 0.6292,  2.8465, -0.1513],
        [-0.8937,  2.1592,  0.8479]])

Addition: syntax 2

print(torch.add(x, y))

Out:

tensor([[-0.2879,  0.6880,  0.2436],
        [ 0.4299, -0.7803, -1.0528],
        [ 1.5577,  2.9404, -0.5684],
        [ 0.6292,  2.8465, -0.1513],
        [-0.8937,  2.1592,  0.8479]])

Addition: providing an output tensor as argument

result = torch.empty(5, 3)
torch.add(x, y, out=result)
print(result)

Out:

tensor([[-0.2879,  0.6880,  0.2436],
        [ 0.4299, -0.7803, -1.0528],
        [ 1.5577,  2.9404, -0.5684],
        [ 0.6292,  2.8465, -0.1513],
        [-0.8937,  2.1592,  0.8479]])

Addition: in-place

# adds x to y
y.add_(x)
print(y)

Out:

tensor([[-0.2879,  0.6880,  0.2436],
        [ 0.4299, -0.7803, -1.0528],
        [ 1.5577,  2.9404, -0.5684],
        [ 0.6292,  2.8465, -0.1513],
        [-0.8937,  2.1592,  0.8479]])

Note

Any operation that mutates a tensor in-place is post-fixed with an _. For example: x.copy_(y), x.t_(), will change x.

You can use standard NumPy-like indexing with all bells and whistles!

print(x[:, 1])

Out:

tensor([-0.2255, -1.5954,  2.5789,  2.2555,  1.6418])

Resizing: If you want to resize/reshape tensor, you can use torch.view:

x = torch.randn(4, 4)
y = x.view(16)
z = x.view(-1, 8)  # the size -1 is inferred from other dimensions
print(x.size(), y.size(), z.size())

Out:

torch.Size([4, 4]) torch.Size([16]) torch.Size([2, 8])

If you have a one element tensor, use .item() to get the value as a Python number

x = torch.randn(1)
print(x)
print(x.item())

Out:

tensor([1.3868])
1.3868496417999268

Read later:

100+ Tensor operations, including transposing, indexing, slicing, mathematical operations, linear algebra, random numbers, etc., are described here.

NumPy Bridge

Converting a Torch Tensor to a NumPy array and vice versa is a breeze.

The Torch Tensor and NumPy array will share their underlying memory locations, and changing one will change the other.

Converting a Torch Tensor to a NumPy Array

a = torch.ones(5)
print(a)

Out:

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

b = a.numpy()
print(b)

Out:

[1. 1. 1. 1. 1.]

See how the numpy array changed in value.

a.add_(1)
print(a)
print(b)

Out:

tensor([2., 2., 2., 2., 2.])
[2. 2. 2. 2. 2.]

Converting NumPy Array to Torch Tensor

See how changing the np array changed the Torch Tensor automatically

import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
np.add(a, 1, out=a)
print(a)
print(b)

Out:

[2. 2. 2. 2. 2.]
tensor([2., 2., 2., 2., 2.], dtype=torch.float64)

All the Tensors on the CPU except a CharTensor support converting to NumPy and back.

CUDA Tensors

Tensors can be moved onto any device using the .to method.

# let us run this cell only if CUDA is available
# We will use ``torch.device`` objects to move tensors in and out of GPU
if torch.cuda.is_available():
    device = torch.device("cuda")          # a CUDA device object
    y = torch.ones_like(x, device=device)  # directly create a tensor on GPU
    # or just use strings ``.to("cuda")``
    x = x.to(device)
    z = x + y
    print(z)
    # ``.to`` can also change dtype together!
    print(z.to("cpu", torch.double))