問題背景:昨天瞭解清除了AlexNet的原理部分,今天着手看看代碼。BTW,我是參考別人的文章,不算原創,不過文中會有很多我自己的見解與補充,算是做個筆記吧。
原文鏈接:
https://blog.csdn.net/taoyanqi8932/article/details/71081390 @yqtaowhu的
首先還是PO一下AlexNet的網絡結構:
@yqtaowhu使用的是CPU運行,怕耗時太久,因此只定義網絡結構,並進行前向後向的測試.
# 此函數爲輸出當前層的參數(層名稱、輸出尺寸)
def print_activations(t): #輸入爲tensor
print(t.op.name, ' ', t.get_shape().as_list())
#此函數爲搭建AlexNet的函數
def inference(images):
"""Build the AlexNet model.
Args:
images: Images Tensor
Returns:
pool5: the last Tensor in the convolutional component of AlexNet. #返回conv_5中池化後的結果
parameters: a list of Tensors corresponding to the weights and biases of the AlexNet model. #返回權重和偏差列表
"""
parameters = []
# conv1
with tf.name_scope('conv1') as scope:# 將scope內生成的Variable自動命名爲conv1/xxx,區分不同卷積層之間的組件。
kernel = tf.Variable(tf.truncated_normal([11, 11, 3, 96],dtype=tf.float32,stddev=1e-1), name='weights') #conv1的核是11*11,3是輸入channel數(後面的函數run_benchmark()中指定了image的channel是3),96是輸出channel數.stddev標準差0.1
conv = tf.nn.conv2d(images, kernel, [1, 4, 4, 1], padding='SAME')
biases = tf.Variable(tf.constant(0.0, shape=[96],dtype=tf.float32),trainable=True, name='biases') #初始化biases,一維向量,96個元素,全爲0
bias = tf.nn.bias_add(conv, biases) #把初始biases加到conv
conv1 = tf.nn.relu(bias, name=scope) #計算relu
print_activations(conv1) #輸出當前層參數
parameters += [kernel, biases] #更新權重和偏差
# lrn1 沒有用LRN層
# TODO(shlens, jiayq): Add a GPU version of local response normalization.
# 考慮到LRN層效果不明顯,而且會讓forward和backwood的速度大大下降所以在此沒有添加
# pool1
pool1 = tf.nn.max_pool(conv1,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1], #各個維度上的移動步長
padding='VALID',
name='pool1')
print_activations(pool1)
# conv2 第二個卷積層,和conv1類似
with tf.name_scope('conv2') as scope:
kernel = tf.Variable(tf.truncated_normal([5, 5, 96, 256], dtype=tf.float32,stddev=1e-1), name='weights')
conv = tf.nn.conv2d(pool1, kernel, [1, 1, 1, 1], padding='SAME')
biases = tf.Variable(tf.constant(0.0, shape=[256], dtype=tf.float32),trainable=True, name='biases')
bias = tf.nn.bias_add(conv, biases)
conv2 = tf.nn.relu(bias, name=scope)
parameters += [kernel, biases]
print_activations(conv2)
# pool2
pool2 = tf.nn.max_pool(conv2,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1],
padding='VALID',
name='pool2')
print_activations(pool2)
# conv3
with tf.name_scope('conv3') as scope:
kernel = tf.Variable(tf.truncated_normal([3, 3, 256, 384],
dtype=tf.float32,
stddev=1e-1), name='weights')
conv = tf.nn.conv2d(pool2, kernel, [1, 1, 1, 1], padding='SAME')
biases = tf.Variable(tf.constant(0.0, shape=[384], dtype=tf.float32),trainable=True, name='biases')
bias = tf.nn.bias_add(conv, biases)
conv3 = tf.nn.relu(bias, name=scope)
parameters += [kernel, biases]
print_activations(conv3)
# conv4
with tf.name_scope('conv4') as scope:
kernel = tf.Variable(tf.truncated_normal([3, 3, 384, 384],dtype=tf.float32,stddev=1e-1),name='weights')
conv = tf.nn.conv2d(conv3, kernel, [1, 1, 1, 1], padding='SAME')
biases = tf.Variable(tf.constant(0.0, shape=[384], dtype=tf.float32),trainable=True, name='biases')
bias = tf.nn.bias_add(conv, biases)
conv4 = tf.nn.relu(bias, name=scope)
parameters += [kernel, biases]
print_activations(conv4)
# conv5
with tf.name_scope('conv5') as scope:
kernel = tf.Variable(tf.truncated_normal([3, 3, 384, 256],dtype=tf.float32,stddev=1e-1),name='weights')
conv = tf.nn.conv2d(conv4, kernel, [1, 1, 1, 1], padding='SAME')
biases = tf.Variable(tf.constant(0.0, shape=[256], dtype=tf.float32),trainable=True, name='biases')
bias = tf.nn.bias_add(conv, biases)
conv5 = tf.nn.relu(bias, name=scope)
parameters += [kernel, biases]
print_activations(conv5)
# pool5
pool5 = tf.nn.max_pool(conv5,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1],
padding='VALID',
name='pool5')
print_activations(pool5)
return pool5, parameters
關於代碼裏的一些函數:
1.tf.truncated_normal() :產生截斷正太分佈。參考鏈接:
https://blog.csdn.net/uestc_c2_403/article/details/72235565
2.tf.nn.conv2d():實現卷積。參考鏈接:
https://www.cnblogs.com/qggg/p/6832342.html
這個鏈接寫得很好很詳細!強烈推薦!
3.關於tensor中的batch.參考鏈接:
http://hp.stuhome.net/index.php/2016/09/20/tensorflow_batch_minibatch/
4.有一個關於部分Tensorflow Python API 的集合:
https://www.jianshu.com/p/e3a79eac554f
5.tf.Variable():定義變量,參考鏈接:
https://blog.csdn.net/gg_18826075157/article/details/78368924
#此函數爲測試函數,image是隨機生成的數據,不是真實的image
def run_benchmark():
"""Run the benchmark on AlexNet."""
with tf.Graph().as_default():
# Generate some dummy images. 生成一些假image
image_size = 224
# Note that our padding definition is slightly different the cuda-convnet.
# In order to force the model to start with the same activations sizes,
# we add 3 to the image_size and employ VALID padding above.
images = tf.Variable(tf.random_normal([FLAGS.batch_size,
image_size,
image_size, 3],
dtype=tf.float32,
stddev=1e-1))
# Build a Graph that computes the logits predictions from the
# inference model.
pool5, parameters = inference(images)
# Build an initialization operation.所有變量初始化
init = tf.global_variables_initializer()
# Start running operations on the Graph.
config = tf.ConfigProto()
config.gpu_options.allocator_type = 'BFC' #BFC算法,防止溢出
sess = tf.Session(config=config)
sess.run(init)
# Run the forward benchmark.
time_tensorflow_run(sess, pool5, "Forward") #前饋,這個函數是自定義的,在後面講解
# Add a simple objective so we can calculate the backward pass.
objective = tf.nn.l2_loss(pool5)
# Compute the gradient with respect to all the parameters.根據梯度更新參數
grad = tf.gradients(objective, parameters)
# Run the backward benchmark.
time_tensorflow_run(sess, grad, "Forward-backward")
關於代碼裏的一些函數:
1.tf.ConfigProto():配置硬件參數,參考鏈接
https://blog.csdn.net/u012436149/article/details/53837651
2.tf.nn.l2_loss() :L2損失函數,參考鏈接:
http://www.cnblogs.com/lovephysics/p/7222488.html
#此函數爲測試程序跑起來用時多久
def time_tensorflow_run(session, target, info_string):
"""Run the computation to obtain the target tensor and print timing stats.
Args:
session: the TensorFlow session to run the computation under.
target: the target Tensor that is passed to the session's run() function.
info_string: a string summarizing this run, to be printed with the stats.
Returns:
None
"""
num_steps_burn_in = 10 #頭幾輪迭代有顯存加載、cache命中等問題因此可以跳過,只考量10輪迭代之後的計算時間
total_duration = 0.0 #總耗時
total_duration_squared = 0.0 #總耗時的平方,用來後面計算方差
for i in xrange(FLAGS.num_batches + num_steps_burn_in):#每一輪迭代
start_time = time.time() #開始時間是當前時間
_ = session.run(target) #跑起來
duration = time.time() - start_time #耗時=當前(結束)時間-開始時間(單位是毫秒)
if i >= num_steps_burn_in: #超過10步
if not i % 10: #是10的倍數
print ('%s: step %d, duration = %.3f' %
(datetime.now(), i - num_steps_burn_in, duration))
total_duration += duration #更新總耗時
total_duration_squared += duration * duration #更新總耗時的平方
mn = total_duration / FLAGS.num_batches # 每輪迭代的平均耗時
vr = total_duration_squared / FLAGS.num_batches - mn * mn #方差,衡量離散程度,即離均值遠不遠
sd = math.sqrt(vr) #標準差
print ('%s: %s across %d steps, %.3f +/- %.3f sec / batch' %
(datetime.now(), info_string, FLAGS.num_batches, mn, sd))
、
在我的電腦上跑這個程序:
各層尺寸,沒有問題
這裏提個小問題:每一層尺寸中最前面的128是什麼?
答:batch_size,這個是在main中指定了的。
可以看到 後者比前者耗時久了好多。。。
另外,對比了一下我和其它博主的結果,發現我的電腦好差勁。。。耗時最久嗚嗚嗚