Overview

This course on PyTorch for deep learning and machine learning aims to teach beginners the fundamentals of PyTorch, a Python-based machine learning framework. By the end of the course, learners will be able to understand the basics of deep learning, work with tensors, create and manipulate data sets, build and train PyTorch models, implement neural network classification, delve into computer vision tasks, and handle custom datasets. The course employs a tutorial-based teaching method using practical examples and coding exercises. It is designed for individuals interested in starting their journey in deep learning and machine learning using PyTorch.

Syllabus

Introduction
0. Welcome and "what is deep learning?"
1. Why use machine/deep learning?
2. The number one rule of ML
3. Machine learning vs deep learning
4. Anatomy of neural networks
5. Different learning paradigms
6. What can deep learning be used for?
7. What is/why PyTorch?
8. What are tensors?
9. Outline
10. How to and how not to approach this course
11. Important resources
12. Getting setup
13. Introduction to tensors
14. Creating tensors
17. Tensor datatypes
18. Tensor attributes information about tensors
19. Manipulating tensors
20. Matrix multiplication
23. Finding the min, max, mean & sum
25. Reshaping, viewing and stacking
26. Squeezing, unsqueezing and permuting
27. Selecting data indexing
28. PyTorch and NumPy
29. Reproducibility
30. Accessing a GPU
31. Setting up device agnostic code
33. Introduction to PyTorch Workflow
34. Getting setup
35. Creating a dataset with linear regression
36. Creating training and test sets the most important concept in ML
38. Creating our first PyTorch model
40. Discussing important model building classes
41. Checking out the internals of our model
42. Making predictions with our model
43. Training a model with PyTorch intuition building
44. Setting up a loss function and optimizer
45. PyTorch training loop intuition
48. Running our training loop epoch by epoch
49. Writing testing loop code
51. Saving/loading a model
54. Putting everything together
60. Introduction to machine learning classification
61. Classification input and outputs
62. Architecture of a classification neural network
64. Turing our data into tensors
66. Coding a neural network for classification data
68. Using torch.nn.Sequential
69. Loss, optimizer and evaluation functions for classification
70. From model logits to prediction probabilities to prediction labels
71. Train and test loops
73. Discussing options to improve a model
76. Creating a straight line dataset
78. Evaluating our model's predictions
79. The missing piece – non-linearity
84. Putting it all together with a multiclass problem
88. Troubleshooting a mutli-class model
92. Introduction to computer vision
93. Computer vision input and outputs
94. What is a convolutional neural network?
95. TorchVision
96. Getting a computer vision dataset
98. Mini-batches
99. Creating DataLoaders
103. Training and testing loops for batched data
105. Running experiments on the GPU
106. Creating a model with non-linear functions
108. Creating a train/test loop
112. Convolutional neural networks overview
113. Coding a CNN
114. Breaking down nn.Conv2d/nn.MaxPool2d
118. Training our first CNN
120. Making predictions on random test samples
121. Plotting our best model predictions
123. Evaluating model predictions with a confusion matrix
126. Introduction to custom datasets
128. Downloading a custom dataset of pizza, steak and sushi images
129. Becoming one with the data
132. Turning images into tensors
136. Creating image DataLoaders
137. Creating a custom dataset class overview
139. Writing a custom dataset class from scratch
142. Turning custom datasets into DataLoaders
143. Data augmentation
144. Building a baseline model
147. Getting a summary of our model with torchinfo
148. Creating training and testing loop functions
151. Plotting model 0 loss curves
152. Overfitting and underfitting
155. Plotting model 1 loss curves
156. Plotting all the loss curves
157. Predicting on custom data