PyTorch Tutorial

PyTorch is an open-source machine learning library mainly used for research and development in fields such as computer vision (CV), natural language processing (NLP), and speech recognition.

PyTorch was developed by Facebook's AI research team and is widely used in the machine learning and deep learning communities.

PyTorch is known for its flexibility and ease of use, making it especially suitable for deep learning research and development.

Who should read this tutorial?

As long as you have basic programming knowledge, you can read this tutorial. Learning PyTorch is suitable for people interested in deep learning and machine learning, including data scientists, engineers, researchers, and students.

What you need to know before reading this tutorial:

Before you start reading this tutorial, the basic knowledge you must have includes Python programming, basic mathematics (linear algebra, probability theory, calculus), basic concepts of machine learning, knowledge of neural networks, and a certain ability to read English to consult documentation and resources.

  • Programming basics: Be familiar with at least one programming language, especiallyPython, because PyTorch is mainly written in Python.

  • Mathematics basics: Understand basic mathematical knowledge such as linear algebra, probability and statistics, and calculus. These are the cornerstones for understanding and implementing machine learning algorithms.

  • Machine learning basics: Understand the basic concepts of machine learning, such as supervised learning, unsupervised learning, reinforcement learning, and model evaluation metrics (accuracy, recall, F1 score, etc.).

  • Deep learning basics: Be familiar with the basic concepts of neural networks, including feedforward neural networks, convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory networks (LSTM), and so on.

  • Computer vision and natural language processing basics: If you plan to apply PyTorch in these areas, having relevant background knowledge will be very helpful.

  • Linux/Unix basics: Although not required, understanding the basics of the Linux/Unix operating system can help you use command-line tools and scripts more effectively, especially in data preprocessing and model training.

  • English reading ability: Since many documents, tutorials, and community discussions are in English, having some English reading ability will help you learn and solve problems more effectively.

Example

The following are some basic tensor operations in PyTorch: how to create random tensors, perform element-wise operations, access specific elements, and compute sums and maximums.

Example

import torch

# Set data type and device
dtype = torch.float  # Tensor data type is floating point
device = torch.device("cpu")  # This computation is performed on CPU

# Create and print two random tensors a and b
a = torch.randn(2, 3, device=device, dtype=dtype)  # Create a random tensor of size 2x3
b = torch.randn(2, 3, device=device, dtype=dtype)  # Create another random tensor of size 2x3

print("Tensor a:")
print(a)

print("Tensor b:")
print(b)

# Multiply element-wise and output the result
print("Element-wise product of a and b:")
print(a * b)

# Output the sum of all elements of tensor a
print("Sum of all elements of tensor a:")
print(a.sum())

# Output the element at row 2, column 3 in tensor a (note that indexing starts from 0)
print("Element at row 2, column 3 of tensor a:")
print(a[1, 2])

# Output the maximum value in tensor a
print("Maximum value in tensor a:")
print(a.max())

Creating tensors:

  • torch.randn(2, 3)Create a tensor with 2 rows and 3 columns, filled with random numbers (following a normal distribution).
  • device=deviceanddtype=dtypeSpecify the computation device (CPU or GPU) and data type (floating point) respectively.

Tensor operations:

  • a * b: Element-wise multiplication.
  • a.sum(): Compute tensorathe sum of all elements.
  • a[1, 2]: Access the tensoraelement at row 2, column 3 (note that indexing starts from 0).
  • a.max(): Get tensoramaximum value in.

Output: (values will differ each run)

张量 a:
tensor([[-0.1460, -0.3490,  0.3705],
        [-1.1141,  0.7661,  1.0823]])

张量 b:
tensor([[ 0.6901, -0.9663,  0.3634],
        [-0.6538, -0.3728, -1.1323]])

a 和 b 的逐元素乘积:
tensor([[-0.1007,  0.3372,  0.1346],
        [ 0.7284, -0.2856, -1.2256]])

张量 a 所有元素的总和:
tensor(0.6097)

张量 a 第 2 行第 3 列的元素:
tensor(1.0823)

张量 a 中的最大值:
tensor(1.0823)

Reference Links

PyTorch official website:https://pytorch.org/

PyTorch official getting-started tutorial:https://pytorch.org/get-started/locally/

PyTorch official documentation:https://pytorch.org/docs/stable/index.html

PyTorch source code:https://github.com/pytorch/pytorch

Other extensions