Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter

System training

To train the agent, we need to create the util/trainer.py file, which provides the train() function. Let's take a look:

1. The signature is as follows:

train(state, nn, filename, args = {})

The function accepts a State class, a neural network class, and other arguments. It also accepts a filename, which is the path of the file containing the generated gameplays. After training, we have the option of saving the output to another model file, as provided in the train() function of the command/train.py file.

2. This command has the following signature:

usage: run.py train [-h]              [--progress, help='show progress bar']              [--epoch EPOCH, help='training epochs']              [--batch BATCH, help='batch size']              [--block BLOCK, help='block size'] [--gpu ...