sensorCar/experimentNetTF.py at master · jcjgraf/sensorCar · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
import tensorflow as tf
import numpy as np
import os
import shutil

from functools import wraps

def callOnce(inputFunc):
	attribute = "_cache_" + inputFunc.__name__

	@property
	@wraps(inputFunc)
	def checkAttribute(self):
		if not hasattr(self, attribute):
			setattr(self, attribute, inputFunc(self))
		return getattr(self, attribute)

	return checkAttribute


class ExperimentNetTF:

	def __init__(self, shape, learningRate):
		self.shape = shape
		self.x = tf.placeholder(tf.float32, shape=[None, self.shape[0]], name="InputData")
		self.y = tf.placeholder(tf.float32, shape=[None, self.shape[-1]], name="LabelData")

		self.weights = self._getInitWeights()

		self.logDir = "./log/experiment"
		shutil.rmtree(self.logDir)
		os.makedirs(self.logDir)


		self.learningRate = learningRate

		self.summaryWriter = tf.summary.FileWriter(self.logDir, graph=tf.get_default_graph())

		self.sess = tf.Session()
		self.sess.run(tf.global_variables_initializer())

		self.output
		self.optimizer
		self.loss

		tf.summary.scalar("loss", self.loss)
		self.mergedSummary = tf.summary.merge_all()

	def _getInitWeights(self):
		return [tf.Variable(tf.truncated_normal([fromLayer, toLayer], stddev=0.1, name="Weight{}".format(i))) for i, (fromLayer, toLayer) in enumerate(zip(self.shape[:-1], self.shape[1:]))]

	def train(self, datasetPath, epochs=1):

		costMeanList = []

		for epoch in range(epochs):
			print(f"Epoch {epoch + 1}")
			with open(datasetPath, "r") as ds:

				costList = []

				for i, line in enumerate(ds):
					lineEntities = np.array([float(i) for i in line.split(",")], dtype=np.float128)

					inputs = np.reshape(lineEntities[:3], (1, 3))
					labels = np.reshape(np.divide(lineEntities[3:], 25), (1, 1))

					# inputs = np.reshape(lineEntities[:2], (1, 2))
					# labels = np.reshape(lineEntities[2:], (1, 1))

					_, loss, summary = self.sess.run([self.optimizer, self.loss, self.mergedSummary], {self.x: inputs, self.y: labels})

					costList.append(loss)
					self.summaryWriter.add_summary(summary, epoch * 1000 + epoch + i)

				tempList = np.array(costList)
				costMeanList.append(np.mean(tempList))

				addListSummary = tf.Summary()
				addListSummary.value.add(tag="MeanLoss", simple_value=np.mean(tempList))
				self.summaryWriter.add_summary(addListSummary, epoch)
				self.summaryWriter.flush()

		self.saveTrainingData("./experimentSave/test.txt", costMeanList)


	def getPrediction(self, xData):
		return self.sess.run(self.output, {self.x: xData})

	@callOnce
	def output(self):
		layerInput = self.x

		for weight in self.weights:
			layerInput = tf.math.tanh(tf.matmul(layerInput, weight))

		return layerInput

	@callOnce
	def loss(self):
		return tf.reduce_mean(tf.square(self.y - self.output))
		# return tf.square(self.y - self.output)

	@callOnce
	def optimizer(self):
		return tf.train.GradientDescentOptimizer(self.learningRate).minimize(self.loss)

	def saveTrainingData(self, filePath, lossList):

		file = open(filePath, "a")

		for loss in lossList:

			file.write(str(loss) + "\n")

		file.close()

	def doSave(self, step):
		savePath = self.saver.save(self.sess, os.path.join(self.savePath, "model"), global_step = step)
		print("Saved current model to {}".format(savePath))


if __name__ == '__main__':
	net = ExperimentNetTF([3, 10, 1], learningRate=0.0005)

	net.train("simulation/dataset/trackMaster1k.txt", epochs=10)

	# net.train("simulation/dataset/testAnd.txt", epochs=100)