我想用TensorFlow以我的方式对mnist的数据集进行分类 首先，将数据集转换为TFRecords文件 然后，使用tf.TFRecodsReader读取此文件 最后，训练数据集。
现有问题：代码没有任何语法错误。但是，我只能在测试数据上获得10％的准确度。
train.py：
import tensorflow as tf
import Net
import os
import numpy as np
import datetime
import time
import tfrecords as rd
BATCH_SIZE = 100
LEARNING_RATE_BASE = 0.01
LEARNING_RATE_DECAY = 0.99
REGULARIZATION_RATE = 0.0001
TRAINING_STEPS = 10000
MOVING_AVERAGE_DECAY = 0.99
MODEL_SAVE_PATH = "Model/"
MODEL_NAME = "model"
def train():
x = tf.placeholder(tf.float32, [
BATCH_SIZE,
Net.IMAGE_SIZE,
Net.IMAGE_SIZE,
Net.NUM_CHANNELS],
name='x-input')
y_ = tf.placeholder(tf.float32, [None, Net.OUTPUT_NODE], name='y-input')
regularizer = tf.contrib.layers.l2_regularizer(REGULARIZATION_RATE)
y = Net.inference(x, True, regularizer)
global_step = tf.Variable(0, trainable=False)
variable_averages = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
variables_averages_op = variable_averages.apply(tf.trainable_variables())
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy)
loss = cross_entropy_mean + tf.add_n(tf.get_collection('losses'))
learning_rate = tf.train.exponential_decay(
LEARNING_RATE_BASE,
global_step,
55000 / BATCH_SIZE, LEARNING_RATE_DECAY,
staircase=True)
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)
with tf.control_dependencies([train_step, variables_averages_op]):
train_op = tf.no_op(name='train')
init_op = tf.group(tf.global_variables_initializer(), tf.local_variables_initializer())
img, one_hot = rd.read_and_decode("./tfrecords/train.tfrecords")
img_batch, label_batch = tf.train.shuffle_batch(tensors = [img, one_hot], batch_size = BATCH_SIZE, capacity = 10000 + 3 * 100, min_after_dequeue = 10000)
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(init_op)
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord = coord, sess = sess)
try:
steps = 1
while not coord.should_stop():
if steps > TRAINING_STEPS:
break
xs, ys = sess.run([img_batch, label_batch])
_, loss_value, step = sess.run([train_op, loss, global_step], feed_dict={x: xs, y_: ys})
if steps % 1000 == 0:
print("After %d training step(s), loss on training batch is %g." % (step, loss_value))
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step=global_step)
steps += 1
except tf.errors.OutOfRangeError:
print("Done training after reading all data")
finally:
coord.request_stop()
coord.join(threads)
sess.close()
def main(argv=None):
start_time = datetime.datetime.now()
print("start_time = "),
print(start_time)
rd.create_record("train")
train_start_time = datetime.datetime.now()
print("train_start_time = " ),
print( train_start_time)
train()
end_time = datetime.datetime.now()
print("end_time = " ),
print(end_time)
if __name__ == '__main__':
main()
Net.py：
import tensorflow as tf
INPUT_NODE = 784
OUTPUT_NODE = 10
IMAGE_SIZE = 28
NUM_CHANNELS = 1
NUM_LABELS = 10
CONV1_DEEP = 32
CONV1_SIZE = 5
CONV2_DEEP = 64
CONV2_SIZE = 5
FC_SIZE = 512
def inference(input_tensor, train, regularizer):
with tf.variable_scope('layer1-conv1'):
conv1_weights = tf.get_variable(
"weight", [CONV1_SIZE, CONV1_SIZE, NUM_CHANNELS, CONV1_DEEP],
initializer=tf.truncated_normal_initializer(stddev=0.1))
conv1_biases = tf.get_variable("bias", [CONV1_DEEP], initializer=tf.constant_initializer(0.0))
conv1 = tf.nn.conv2d(input_tensor, conv1_weights, strides=[1, 1, 1, 1], padding='SAME')
relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))
with tf.name_scope("layer2-pool1"):
pool1 = tf.nn.max_pool(relu1, ksize = [1,2,2,1],strides=[1,2,2,1],padding="SAME")
with tf.variable_scope("layer3-conv2"):
conv2_weights = tf.get_variable(
"weight", [CONV2_SIZE, CONV2_SIZE, CONV1_DEEP, CONV2_DEEP],
initializer=tf.truncated_normal_initializer(stddev=0.1))
conv2_biases = tf.get_variable("bias", [CONV2_DEEP], initializer=tf.constant_initializer(0.0))
conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding='SAME')
relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_biases))
with tf.name_scope("layer4-pool2"):
pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
pool_shape = pool2.get_shape().as_list()
nodes = pool_shape[1] * pool_shape[2] * pool_shape[3]
reshaped = tf.reshape(pool2, [pool_shape[0], nodes])
with tf.variable_scope('layer5-fc1'):
fc1_weights = tf.get_variable("weight", [nodes, FC_SIZE],
initializer=tf.truncated_normal_initializer(stddev=0.1))
if regularizer != None: tf.add_to_collection('losses', regularizer(fc1_weights))
fc1_biases = tf.get_variable("bias", [FC_SIZE], initializer=tf.constant_initializer(0.1))
fc1 = tf.nn.relu(tf.matmul(reshaped, fc1_weights) + fc1_biases)
if train: fc1 = tf.nn.dropout(fc1, 0.5)
with tf.variable_scope('layer6-fc2'):
fc2_weights = tf.get_variable("weight", [FC_SIZE, NUM_LABELS],
initializer=tf.truncated_normal_initializer(stddev=0.1))
if regularizer != None: tf.add_to_collection('losses', regularizer(fc2_weights))
fc2_biases = tf.get_variable("bias", [NUM_LABELS], initializer=tf.constant_initializer(0.1))
logit = tf.matmul(fc1, fc2_weights) + fc2_biases
return logit
tfrecords.py：
import os
import tensorflow as tf
from PIL import Image
import Net
def create_record(op_type):
writer = tf.python_io.TFRecordWriter("./tfrecords/" + op_type + ".tfrecords")
f = open("./" + op_type + ".txt", 'r')
img_num = 0
for line in f.readlines():
img_num += 1
if img_num % 2000 == 0:
print("already read in %d images." % (img_num))
str_split = line.split()
img_path = "./" + str_split[0]
index = int(str_split[1])
img = Image.open(img_path)
img = img.resize((Net.IMAGE_SIZE, Net.IMAGE_SIZE))
img_raw = img.tobytes()
example = tf.train.Example(features=tf.train.Features(feature={
'label': tf.train.Feature(int64_list=tf.train.Int64List(value=[index])),
'img_raw': tf.train.Feature(bytes_list=tf.train.BytesList(value=[img_raw]))
}))
writer.write(example.SerializeToString())
writer.close()
def read_and_decode(filename):
filename_queue = tf.train.string_input_producer([filename], shuffle = True)
reader = tf.TFRecordReader()
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(serialized_example,
features={
'label': tf.FixedLenFeature([], tf.int64),
'img_raw' : tf.FixedLenFeature([], tf.string),
})
img = tf.decode_raw(features['img_raw'], tf.uint8)
img = tf.reshape(img, [Net.IMAGE_SIZE, Net.IMAGE_SIZE, Net.NUM_CHANNELS])
img = 1 - tf.cast(img, tf.float32) * (1. / 255)
label = tf.cast(features['label'], tf.int32)
one_hot = tf.one_hot(label, 10, dtype = tf.float32)
one_hot = tf.reshape(one_hot, [-1])
return img, one_hot
Eval.py：
import time
import math
import tensorflow as tf
import numpy as np
import Net
import Train
import tfrecords as rd
def evaluate():
with tf.Graph().as_default() as g:
x = tf.placeholder(tf.float32, [
10000,
Net.IMAGE_SIZE,
Net.IMAGE_SIZE,
Net.NUM_CHANNELS],
name='x-input')
y_ = tf.placeholder(tf.float32, [None, Net.OUTPUT_NODE], name='y-input')
#validate_feed = {x: mnist.test.images, y_: mnist.test.labels}
global_step = tf.Variable(0, trainable=False)
regularizer = tf.contrib.layers.l2_regularizer(Train.REGULARIZATION_RATE)
y = Net.inference(x, False, regularizer)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
variable_averages = tf.train.ExponentialMovingAverage(Train.MOVING_AVERAGE_DECAY)
variables_to_restore = variable_averages.variables_to_restore()
saver = tf.train.Saver(variables_to_restore)
for i in range(1):
img, one_hot = rd.read_and_decode("./tfrecords/test.tfrecords")
img_batch, label_batch = tf.train.shuffle_batch(tensors = [img, one_hot], batch_size = 10000, capacity = 10000 + 3 * 100, min_after_dequeue = 10000)
with tf.Session() as sess:
ckpt = tf.train.get_checkpoint_state(Train.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
#the reason of this error!!!
#tf.global_variables_initializer().run()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(coord = coord, sess = sess)
try:
xs, ys = sess.run([img_batch, label_batch])
accuracy_score = sess.run(accuracy, feed_dict={x:xs, y_:ys})
print("After %s training step(s), test accuracy = %g" % (global_step, accuracy_score))
except tf.errors.OutOfRangeError:
print("Done testting after reading all data")
finally:
coord.request_stop()
coord.join(threads)
sess.close()
else:
print('No checkpoint file found')
return
def main(argv=None):
rd.create_record("test")
evaluate()
if __name__ == '__main__':
main()
现在，我已经解决了这个问题。
我的代码和其他flie：https://github.com/xmy7216/MNIST_classification.git
Linux：Red Hat Enterprise Linux Server 7.2（Maipo）
GPU：特斯拉P4
TensorFlow：1.3
的Python：2.7.5
答案 0 :(得分：0)
我找到原因。因为在使用saver.restore加载模型后，我再次按tf.global_variables_initializer().run()初始化变量。
我多么愚蠢！