Resources: PyTorch

Topics to Look For¶

• Tensors
• Autograd
• Defining a network
• Forward and backprop
• Optimization algorithms
• Training
• Testing
• CUDA/GPU execution

Resources¶

• I’ve made some small improvements to the PyTorch “60 Minute Blitz” tutorial. It is a series of four notebooks:
  • PyTorch Tutorial 1: Tensors
  • PyTorch Tutorial 2: Autograd
  • PyTorch Tutorial 3: Neural Networks
  • PyTorch Tutorial 4: A Full Example

Read through these carefully. They will not help much if you just read them. To learn from these resources you need to read each paragraph, think about it to ensure you have some sense of what it’s saying, and (very importantly) read the example code and look at its output, if any, with the goal of figuring out what that example code is trying to teach you.

And of course, as Jupyter notebooks, they can be modified and re-executed to dig into the code and concepts more yourself.