Skip to content

Lab 4: Transformers

Video

Lab by Sergey Karayev.

In this lab, you use the LineCNN + LSTM model with CTC loss from lab 3 as an "encoder" of the image, and then send it through Transformer decoder layers.

  • 00:00 - Introduction
  • 01:43 - LineCNNTransformer class
  • 04:50 - TransformerLitModel
  • 06:51 - Code to make predictions
  • 08:50 - Training guidelines

Follow along

Readme