The following documents, use tensorflow started to learn the most basic example mnist, the most intuitive and easiest way to use shows tensorboard
Wherein the tensorboard comprising: scalar, image, histogram, and wherein the code space down to display.
tensorboard_test.py
#coding:. 8 UTF-
Import tensorflow TF AS
from tensorflow.examples.tutorials.mnist Import Input_Data
Import Time
Import OS
Import PROJECTOR_visual
"" "
weights initialized
initialized to a small positive number close to 0
" ""
DEF weight_variable (Shape) :
Initial = tf.truncated_normal (Shape, STDDEV = 0.1)
return tf.Variable (Initial)
DEF bias_variable (Shape):
Initial tf.constant = (0.1, Shape = Shape)
return tf.Variable (Initial)
"" "
convolution and pooling of the convolution steps of 1 (stride size), 0 margins (padding size)
pooling max pooling do traditional 2x2 size with a simple template
"" "
DEF conv2d (X, W is):
return TF. nn.conv2d (x, W, strides = [1, 1, 1, 1],padding = 'SAME')
# tf.nn.conv2d(input, filter, strides, padding, use_cudnn_on_gpu=None, data_format=None, name=None)
# x(input) : [batch, in_height, in_width, in_channels]
# W(filter) : [filter_height, filter_width, in_channels, out_channels]
# strides : The stride of the sliding window for each dimension of input.
# For the most common case of the same horizontal and vertices strides, strides = [1, stride, stride, 1]
def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize = [1, 2, 2, 1],
strides = [1, 2, 2, 1], padding = 'SAME')
# tf.nn.max_pool(value, ksize, strides, padding, data_format='NHWC', name=None)
# x(value) : [batch, height, width, channels]
# Ksize (pool size): A List of INTS that has length> = 4. The window of The size for each of Dimension of Tensor The INPUT.
# Strides (sliding the pool size): A list of ints that has length> = 4. The window of a stride of Sliding of The for each of Dimension The iNPUT Tensor.
start = time.clock () # calculation start time
mnist = input_data.read_data_sets ( "MNIST_data /" , one_hot = True) #MNIST data input
os.system ( "python create_sprite.py ")
" ""
The first layer convolutional layer
x_image (BATCH, 28, 28,. 1) -> h_pool1 (BATCH, 14, 14, 32)
"" "
X = tf.placeholder (tf.float32, [ none, 784])
x_image = -1, 28, 28,. 1]) # a dimension representing the number of the last channel tf.reshape (x, [, if compared rgb. 3
tf.summary.image ( 'input_image', x_image,10)
W_conv1 = weight_variable([5, 5, 1, 32])10)
b_conv1 = bias_variable([32])
tf.summary.histogram('W_conv1',W_conv1)
tf.summary.histogram('b_conv1',b_conv1)
conv1 = conv2d(x_image, W_conv1) + b_conv1
h_conv1 = tf.nn.relu(conv1)
# x_image -> [batch, in_height, in_width, in_channels]
# [batch, 28, 28, 1]
# W_conv1 -> [filter_height, filter_width, in_channels, out_channels]
# [5, 5, 1, 32]
# output -> [batch, out_height, out_width, out_channels]
# [batch, 28, 28, 32]
h_pool1 = max_pool_2x2(h_conv1)
# h_conv1 -> [batch, in_height, in_weight, in_channels]
# [batch, 28, 28, 32]
# output -> [batch, out_height, out_weight, out_channels]
# [BATCH, 14, 14, 32]
"" "
The second layer convolutional layer
," ""
h_pool1 (BATCH, 14, 14, 32) -> h_pool2 (BATCH,. 7,. 7, 64)
"" "
W_conv2 weight_variable = ([. 5,. 5, 32, 64])
b_conv2 bias_variable = ([64])
tf.summary .histogram ( 'W_conv2', W_conv2)
tf.summary.histogram ( 'b_conv2', b_conv2)
CONV2 = conv2d (h_pool1, W_conv2) + b_conv2
h_conv2 = tf.nn.relu (CONV2)
# h_pool1 -> [BATCH, 14, 14, 32]
# W_conv2 -> [. 5,. 5, 32, 64]
# Output -> [BATCH, 14, 14, 64]
h_pool2 = max_pool_2x2 (h_conv2)
# h_conv2 -> [BATCH, 14, 14, 64]
# output -> [BATCH,. 7,. 7, 64]
"" "
deconvolution layer, added to the output image
reverse_weight1 = weight_variable ([5,5,32,64])
reverse_conv1 = tf.nn.conv2d_transpose(conv2,reverse_weight1,[50,14,14,32],strides=[1,1,1,1],padding="SAME")
reverse_weight2 = weight_variable([5,5,1,32])
reverse_conv2 = tf.nn.conv2d_transpose(reverse_conv1,reverse_weight2,[50,28,28,1],strides=[1,2,2,1],padding="SAME")
reverse_weight3 = weight_variable([5,5,1,32])
reverse_conv3 = tf.nn.conv2d_transpose(conv1,reverse_weight3,[50,28,28,1],strides=[1,1,1,1],padding="SAME")
tf.summary.image("reverse_conv2",reverse_conv2,10)
tf.summary.image("reverse_conv1",reverse_conv3,10)
"""
第三层 全连接层
h_pool2(batch, 7, 7, 64) -> h_fc1(1, 1024)
"""
W_fc1 = weight_variable([7 * 7 * 64, 1024])
= bias_variable b_fc1 ([1024])feed_dict added control parameters keep_prob dropout ratio
"" "
tf.summary.histogram ( 'W_fc1', W_fc1)
tf.summary.histogram ( 'b_fc1', b_fc1)
h_pool2_flat = tf.reshape (h_pool2, [-1,. 7. 7 * 64 *])
h_fc1 = tf.nn.relu (tf.matmul (h_pool2_flat, W_fc1) + b_fc1)
"" "
Dropout
h_fc1 -> h_fc1_drop, training is enabled, the test close
" ""
keep_prob = tf.placeholder ( "float")
h_fc1_drop = tf.nn.dropout (h_fc1 , keep_prob)
"" "
the fourth layer Softmax output layer
" ""
W_fc2 weight_variable = ([1024, 10])
b_fc2 bias_variable = ([10])
y_conv = tf.nn.softmax (tf.matmul (h_fc1_drop, W_fc2) + b_fc2)
"" "
training and evaluation model
ADAM optimizer do steepest descent gradient,feed_dict added dropout proportional control parameter keep_prob
y_ = tf.placeholder ( "float", [None, 10])
cross_entropy = -tf.reduce_sum (y_ * tf.log ( y_conv)) # calculating cross entropy
train_step = tf.train.AdamOptimizer (1e-4) .minimize (cross_entropy) # adam optimizer to use to learn to perform rate 0.0001 trimming
correct_prediction = tf.equal (tf.argmax (y_conv, 1), tf.argmax (y_, 1)) # tag and determines the prediction matches the actual label
accuracy = tf.reduce_mean (tf.cast (correct_prediction, "float") )
tf.summary.scalar ( 'cross_entropy', cross_entropy)
tf.summary.scalar ( 'Accuracy', Accuracy)
Merged = tf.summary.merge_all ()
sess = tf.Session () # boot models created
writer = tf. summary.FileWriter ( "logs /", sess.graph)
# sess.run (tf.initialize_all_variables ()) # old version
sess.run (tf.global_variables_initializer ()) # initialize variables
for i in range (PROJECTOR_visual.TRAINING_STEPS): # start training model 5000 circuit training
batch = mnist.train.next_batch (50) #batch size to 50
IF I == 0% 100:
train_accuracy = accuracy.eval (the session Sess =,
feed_dict = {X: BATCH [0], Y_: BATCH [. 1], keep_prob: 1.0})
Print ( "% STEP D, G train_accuracy%"% (I, train_accuracy))
sess.run (train_step, feed_dict {X =: BATCH [0], Y_: BATCH [. 1],
keep_prob: 0.5}) # neuron output probability remains unchanged as keep_prob 0.5
RS, sess.run _ = ([Merged, train_step], = {feed_dict X: BATCH [0], Y_: BATCH [. 1], keep_prob: 1.0})
writer.add_summary (RS, I)
final_result = sess.run (h_fc1, feed_dict = {X: mnist.test.images})
Print ( "G% Accuracy Test"% accuracy.eval (= Sess the session,
feed_dict = {X: mnist.test.images, Y_: mnist.test.labels,
keep_prob: 1.0})) # neuron output remains unchanged keep_prob probability is 1, that is unchanged, has maintained output
PROJECTOR_visual.visualisation (final_result)
PPROJECTOR_visual.py
tensorflow TF AS Import
Import OS
Import tqdm
from tensorflow.contrib.tensorboard.plugins Import Projector
TRAINING_STEPS = 1000
. LOG_DIR = 'logs'
SPRITE_FILE = 'mnist_sprite.jpg'
META_FIEL = "mnist_meta.tsv"
TENSOR_NAME = "FINAL_LOGITS"
# Visualizing the final output layers vector required log file
DEF visualization (final_result):
# using a new variable to hold the result of the final output layer vector, because the embedding is by Tensorflow variables completed, so PROJECTOR visualization are TensorFlow Variational wow.
# So here needs a new definition of a variable to hold the value of output layer vector
y_visual = tf.Variable (final_result, name = TENSOR_NAME)
summary_writer tf.summary.FileWriter = (. LOG_DIR)
# by project.ProjectorConfig classes to help generate a log file
config = projector.ProjectorConfig ()
# A need to increase the visual bedding results
embedding config.embeddings.add = ()
# Tensorflow specify the variable name corresponding to the result of embedding
embedding.tensor_name = y_visual.name
# Find the Specify WHERE you The Metadata
# specified embedding result corresponding original Data information. For example, here is the categories specified real test each one corresponding to the image MNIST. In a word may be a word vector corresponding to the ID word.
# This file is optional, if the vector is not specified then no label.
= META_FIEL embedding.metadata_path
# Find the Specify WHERE you The sprite (WE Will Create the this later)
# specified sprite image. This is optional, if no sprite image, the visualization of results
# Each point is a sleepy little spot, rather than a specific picture.
= SPRITE_FILE embedding.sprite.image_path
# when providing sprite images, you can specify a single picture by single_image_dim.
# This will be used to intercept the correct picture from the original sprite image.
embedding.sprite.single_image_dim.extend ([28, 28])
# Say that you want to Visualize The embeddings
# The content PROJECTOR require written to the log file.
projector.visualize_embeddings (summary_writer, config)
log information generated # session initialization newly declared variable and require written to the file.
= tf.InteractiveSession Sess ()
sess.run (tf.global_variables_initializer ())
Saver = tf.train.Saver ()
saver.save (Sess, the os.path.join (. LOG_DIR, "Model"), TRAINING_STEPS)
summary_writer.close ()
The two next .py file in the same folder, and then run when the direct use cmd, execute python tensorboard_test.py, you will be able to start.
Then enter in your browser: http: // localhost: 8080 screen can be opened tensorboard