cruve = np.mean(
[reward_rec[i:i + 10] for i in range(0, len(reward_rec), 10)],
axis=1)
plt.plot(range(len(cruve)), cruve, c='b')
plt.xlabel(r'Epochs $ \times 10$')
plt.ylabel('Rewards')
# Save model
critic_model.save_weights(r'./saved_model/DRL/crtic_3000.HDF5')
print('model saved.')
elif MODE == 'eval':
# Load model
critic_model = Network(num_actions, embeddings_matrix)
_ = critic_model(inputs=[inputs, act_input])
critic_model.load_weights(r'./saved_model/DRL/crtic_3000.HDF5')
print('model loaded.')
actor_q_model = tf.keras.Model(
inputs=critic_model.input,
outputs=critic_model.get_layer('q_outputs').output)
ones = np.ones(shape=(len(test_seg), num_actions))
pred = predict(actor_q_model, (input_validation, ones))
print('On validate data: ')
metrics(y_test, pred)
ones_tr = np.ones(shape=(len(train_seg), num_actions))
pred_tr = predict(actor_q_model, (input_train, ones_tr))
print('On train data: ')
metrics(y_train, pred_tr)
