Building Your First Network in PyTorch

A summary to kickstart your deep learning career.

Photo by Max Duzij on Unsplash

Starting a deep learning project sounds scary and difficult? I have read through articles, took lessons, and watched videos about neural networks, but how do I begin programming one? We have all been through that stage, and this is why I am creating this article to tell you everything (or at least most of the things I know) to begin your PyTorch model training project.

The guide is presented in a bottom-up way. I will first describe individual components that are important to training a deep network, then provide examples on how to combine all the components together for training and testing.

Side Note:

The article serves as a bridging medium for converting the theoretical knowledge in ML directly into codes. Prior ML knowledge is assumed.

My projects are mainly in the domain of computer vision, and so what I found to be the most useful functions in PyTorch are also biased towards applications regarding images.

Table of Contents

· Table of Contents
· Import
· Network Components
∘ Fully-Connected Layers
∘ The Convolution Family
∘ Recurrent Networks
∘ Activation functions
· Loss Functions
· Optimisers
· Setting up GPU
· Combining Everything
∘ Import PyTorch
∘ Create Network
∘ Create Dataset and Dataloader
∘ Initialise Networks, Optimisers, and Schedulers
∘ Training, Evaluating, and Saving
· Pretrained Models
· Other Fun PyTorch Functions
· End Note

Import

There are a lot of libraries associated with PyTorch, here I listed out the essential libraries to perform everything underneath:

import torch
import torch.nn as nn
import torch.optim as optim
import torch.utils.data
from…