Natural Language Processing with Python Quick Start Guide

Putting the pieces together again

These are the pieces which we looked at, let's quickly summarize them:

• Loss function: Binary cross entropy with Logit loss. It serves as the quality metric of how far the predictions are from the ground truth.
  • Optimizer: We use the Adam optimizer with default parameters, set with a learning rate of 1e-2 or 0.01:

This is how we would see these 2 components in PyTorch:

from torch import optimopt = optim.Adam(model.parameters(), lr=1e-2)loss_func = nn.BCEWithLogitsLoss().cuda()

We call set the number of epochs for which the model has to be trained here:

epochs = 3

This is set to a very small value because this entire notebook, model, and training loop is just for demonstrative purposes.