Hands-On Python Deep Learning for the Web

Training the CNTK model

With the model created, we can now move on to training the model and making it learn to predict. To do so, we need to use the CNTK model object and fit the samples in the dataset to it. We can, at the same time, log loss and other evaluation metrics. We need to carry out the following steps to train our model:

1. Create placeholders for loss and the classification error:

loss = C.cross_entropy_with_softmax(model, label)label_error = C.classification_error(model, label)

2. Now, we can set up a trainer object for the CNTK framework, which is used to perform the actual training:

lrs = C.learning_rate_schedule(learning_rate, C.UnitType.minibatch)learner = C.sgd(model.parameters, lrs)trainer = C.Trainer(model, (loss, label_error), ...