c99295e1903952b3b806c4fd641e28f852b17c36,autokeras/utils.py,ModelTrainer,train_model,#ModelTrainer#,101

Before Change

        Train the model with dataset and return the minimum_loss
        self.training_losses = []
        self._no_improvement_count = 0
        self.minimum_loss = float("inf")
        batch_size = min(self.x_train.shape[0], 200)
        if constant.DATA_AUGMENTATION:
            flow = self.datagen.flow(self.x_train, self.y_train, batch_size)
        else:
            flow = None
        for _ in range(constant.MAX_ITER_NUM):
            if constant.DATA_AUGMENTATION:
                self.model.fit_generator(flow, epochs=constant.EPOCHS_EACH)
            else:
                self.model.fit(self.x_train, self.y_train,
                               batch_size=batch_size,
                               epochs=constant.EPOCHS_EACH,
                               verbose=self.verbose)
            loss, _ = self.model.evaluate(self.x_test, self.y_test, verbose=self.verbose)
            if self._converged(loss):
                break
        return self.minimum_loss


def extract_config(network):
    Return configuration of one model

After Change

            config.gpu_options.allow_growth = True
            sess = tf.Session(config=config)
            backend.set_session(sess)
        try:
            if constant.DATA_AUGMENTATION:
                flow = self.datagen.flow(self.x_train, self.y_train, batch_size)
                self.model.fit_generator(flow,
                                         epochs=constant.MAX_ITER_NUM,
                                         validation_data=(self.x_test, self.y_test),
                                         callbacks=callbacks,
                                         verbose=self.verbose)
            else:
                self.model.fit(self.x_train, self.y_train,
                               batch_size=batch_size,
                               epochs=constant.MAX_ITER_NUM,
                               validation_data=(self.x_test, self.y_test),
                               callbacks=callbacks,
                               verbose=self.verbose)
        except NoImprovementError as e:
            if self.verbose:
                print("Training finished!")
                print(e.message)


def extract_config(network):
    Return configuration of one model
    return network.get_config()

Instances