Function Run

Sources/templates/classifier_template.py:22–65

(self)

save_path = ""

def Run(self):

	### defining the model: ###

	#inputs: 
	self.X = tf.placeholder(tf.float32, shape=[None, self._input.input_size], name="x_input")        
	self.y = tf.placeholder(tf.float32, shape=[None, self._input.labels_size], name="y_labels")
   	
    #	ouput: 
	output = self._model.Run(self.X)

	#Loss function
	_loss = tf.reduce_mean(tf.pow(output-self.y, 2))

	#measure the accuracy
	correct_pred = tf.equal(tf.argmax(output, 1), tf.argmax(self.y, 1))
	_accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

	#run the optimizer
	_solver = tf.train.AdamOptimizer(self.learning_rate).minimize(_loss)

	#initalize everything
	instance = AIBlocks.InitModel(load_path=self.save_path)
	Log("Model initialized!")

	#set the progress bar state
	SetState(self.id, 0.001)
        
	for it in range(self.training_iterations):
		batch = self._input.getNextBatch()
            
		X_batch  = batch[0]
		Y_batch  = batch[1]
        
		_, loss, acc = instance.Run([_solver, _loss, _accuracy], feed_dict={self.X: X_batch, self.y: Y_batch})

		#every N steps, send the state to the scene
		if it % self.display_step == 0:
			SetState(self.id, it/self.training_iterations)
			SendChartData(self.id, "Loss", loss, "#ff0000")
			SendChartData(self.id, "Accuracy", acc)
        
	AIBlocks.SaveModel(instance)
	AIBlocks.CloseInstance(instance)