Deep Learning with PyTorch
Learn deep learning in PyTorch, from tensors and gradient descent to image classifiers, residual networks and GANs that generate new images.
Start learning Open course bookAbout this course
PyTorch is the most widely used framework in deep learning research and a popular choice in industry, because models are written as ordinary Python and every step of training is visible. Learning it means learning tensors, automatic differentiation, the training loop and the standard architectures for images and generation.
This course follows a long beginner's course by Daniel Bourke that covers the PyTorch workflow from fundamentals to custom datasets, and then the six-part Deep Learning with PyTorch live course by Aakash N S. The live course goes from tensors and linear regression through logistic regression on images, feedforward networks on GPUs and convolutional networks to residual networks with modern training tricks, and ends with generative adversarial networks. Each lesson comes with an original book chapter of worked examples.
What you will learn
- Create and manipulate tensors and predict the shapes of operations
- Write the PyTorch training and evaluation loop from memory
- Compute gradients, softmax and cross-entropy by hand and check them with autograd
- Build feedforward, convolutional and residual networks and train them on a GPU
- Explain how a generator and discriminator are trained against each other
Who this course is for
Python programmers with some machine learning background who want to build deep learning models. Comfort with NumPy helps; the chapters explain the maths they use.