First we need to create the TensorFlow checkpoint files – just uncomment the two lines for both player1 and player2 in funcs.py and run python play.py again:
if player1version > 0: player1_network = player1_NN.read(env.name, run_version, player1version) player1_NN.model.set_weights(player1_network.get_weights()) # saver = tf.train.Saver() # saver.save(K.get_session(), '/tmp/alphazero19.ckpt') if player2version > 0: player2_network = player2_NN.read(env.name, run_version, player2version) player2_NN.model.set_weights(player2_network.get_weights()) # saver = tf.train.Saver() # saver.save(K.get_session(), '/tmp/alphazero_4.ckpt')
This may look familiar to you as we did something similar in Chapter 8, Predicting Stock Price ...
