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