[TensorFlow] 기초 개념 및 수행 결과 _ 5. Tensorboard : Graph Visualization 사용하기

[TensorFlow] 기초 개념 및 수행 결과 _ 5. Tensorboard : Graph Visualization 사용하기
카테고리 MachineLearning
제목 [TensorFlow] 기초 개념 및 수행 결과 _ 5. Tensorboard : Graph Visualization 사용하기
작성시간 2018-03-29 17:02:59 +0900
조회수 698

- 학습 5은 학습 4를 기반으로 진행합니다

Tensorboard : Graph Visualization 사용하기!!


출처 

In [1]:
import tensorflow as tf
In [2]:
input_data = [[1,5,3,7,8,10,12],
              [5,8,10,3,9,7,1]]
label_data = [[0,0,0,1,0],
              [1,0,0,0,0]]
INPUT_SIZE = 7 
HIDDEN_SIZE_1 = 10
HIDDEN_SIZE_2 = 8
CLASSES = 5
In [3]:
x = tf.placeholder(tf.float32, shape = [None, INPUT_SIZE], name="x" )
y = tf.placeholder(tf.float32, shape = [None, CLASSES], name="y" )
In [4]:
tensor_map = {x : input_data, y : label_data}
In [5]:
weight_hidden_1 = tf.Variable( tf.truncated_normal(shape=[INPUT_SIZE, HIDDEN_SIZE_1]) , dtype=tf.float32, name="weight_hidden_1" ) 
bias_hidden_1 = tf.Variable( tf.zeros(shape=[HIDDEN_SIZE_1]) , dtype=tf.float32, name = "bias_hidden_1" )
weight_hidden_2 = tf.Variable( tf.truncated_normal(shape=[HIDDEN_SIZE_1, HIDDEN_SIZE_2]) , dtype=tf.float32, name="weight_hidden_2" ) 
bias_hidden_2 = tf.Variable( tf.zeros(shape=[HIDDEN_SIZE_2]) , dtype=tf.float32, name = "bias_hidden_2" )
weight_output = tf.Variable( tf.truncated_normal(shape=[HIDDEN_SIZE_2, CLASSES]) , dtype=tf.float32, name="weight_output" ) 
bias_output = tf.Variable( tf.zeros(shape=[CLASSES]) , dtype=tf.float32, name = "bias_output" )
In [6]:
param_list = [weight_hidden_1, bias_hidden_1, weight_hidden_2, bias_hidden_2, weight_output, bias_output ]
saver = tf.train.Saver(param_list)

각각의 scope 에 넣어준다.

In [7]:
with tf.name_scope('hidden_layer_1') as h1scope:
    hidden_1 = tf.matmul( x , weight_hidden_1 , name="hidden_1" ) + bias_hidden_1
with tf.name_scope('hidden_layer_2') as h2scope:
    hidden_2 = tf.matmul( hidden_1 , weight_hidden_2, name="hidden_2" ) + bias_hidden_2
with tf.name_scope('output_layer') as osscope:
    answer = tf.matmul( hidden_2 , weight_output, name="answer" ) + bias_output

일단 accuracy 는 무시하고 ./summaires 디렉토리에 생기는 파일을 주목해보자 !! (acc 가 이 코드는 자꾸 0 이 나온다)

In [9]:
Learning_Rate = 0.05
cost_ = -y*tf.log(answer)-(1-y)*tf.log((1-answer))
cost = tf.reduce_sum(cost_, reduction_indices=1)
cost = tf.reduce_mean(cost, reduction_indices=0)
train = tf.train.GradientDescentOptimizer(Learning_Rate).minimize(cost)
comp_pred = tf.equal( tf.arg_max(answer, 1), tf.arg_max(y, 1) )
accuracy = tf.reduce_mean(tf.cast(comp_pred, tf.float32))
sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init) 
saver.restore(sess, './tensorflow_live.ckpt')
merge = tf.summary.merge_all()
for i in range(1000):
    _ , loss, acc = sess.run([train, cost, accuracy], feed_dict = tensor_map)
    if i%100 == 0:
        train_writer = tf.summary.FileWriter('./summaries', sess.graph)
        saver.save(sess, './tensorflow_live.ckpt')
        print ( '---------------' ) 
        print ( 'step    : %f' %  i  )
        print ( 'loss    : %f' % loss)
        print ( 'accuracy: %f' % acc ) 
    
sess.close()
---------------
step    : 0.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 100.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 200.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 300.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 400.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 500.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 600.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 700.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 800.000000
loss    : nan
accuracy: 0.000000
---------------
step    : 900.000000
loss    : nan
accuracy: 0.000000


tensorboard --logdir=./

summaries 폴더를 현재 디렉토리로 지정하고, 터미널에서 위의 명령어를 사용하여 텐서보드를 활성화시킨다.

텐서 보드를 이용하여 보면 아래와 같은 구조를 가진다







▼discuss 댓글▼



▼facebook 댓글▼