TensorFlow分布式部署【单机多卡】

让TensorFlow飞一会儿

面对大型的深度神经网络训练工程,训练的时间非常重要。训练的时间长短依赖于计算处理器也就是GPU,然而单个GPU的计算能力有限,利用多个GPU进行分布式部署,同时完成一个训练任务是一个很好的办法。对于caffe来说,由于NCCL的存在,可以直接在slover中指定使用的GPU。然而对于Tensorflow,虽然Contrib库中有NCCL,但是我并没有找到相关的例子,所以,还是靠双手成就梦想。

原理简介

TensorFlow支持指定相应的设备来完成相应的操作,所以如何分配任务是很关键的一环。GPU擅长大量计算,所以整个Inference和梯度的计算就交给GPU来做,更新参数的小事情就交给CPU来做。这就比如校长要知道整个年级的平均成绩,就把改卷子的任务分配给每个班的老师,每个班的老师批改完卷子以后,把各自班级的成绩上交给校长,校长计算个平均数就行。在这里,校长就是CPU,每个班级的老师就是GPU。

下面放出一张图来说明问题。

我们可以清楚的看到CPU中保存变量,GPU们计算整个model和gradients,然后把得到的梯度送回CPU中,CPU计算各个GPU送回来梯度的平均值作为本次step的梯度对参数进行更新。从图中我们可以看到只有当所有的GPU完成梯度计算以后,CPU才能求平均值,所以,整个神经网络的迭代速度将取决于最慢的一个GPU,这也就是同步更新。那能不能异步更新呢?当然是可以的把更新参数这个操作也放回到GPU上,但是异步更新会造成训练不稳定,有的快有的慢,你说到底听谁的…

在上图中我们可以看到有几个关键点需要注意:

  1. CPU上定义变量
  2. GPU上分别定义model和gradients操作,得到每个GPU中的梯度
  3. 又回到CPU中计算平均平均梯度,并进行参数更新

Talk is cheap, show me the code!!

好,下面放代码。

示例代码

示例代码分如下几个部分:

  1. 读入数据
  2. 在cpu中定义变量
  3. 搭建Inference
  4. 定义loss
  5. 定义训练过程

读入数据

由于是在不同的GPU上进行运算,所以我们采用TF官方的数据格式tfrecords作为输入,tfrecords的MNIST数据集格式可以在网上很轻易的找到。读入数据的时候我们就用标准的tfrecords数据集读入的格式。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
def read_and_decode(filename_queue):
reader = tf.TFRecordReader()
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(
serialized_example,
# Defaults are not specified since both keys are required.
features={
'image_raw': tf.FixedLenFeature([], tf.string),
'label': tf.FixedLenFeature([], tf.int64),
})

image = tf.decode_raw(features['image_raw'], tf.uint8)
image.set_shape([IMAGE_PIXELS])
image = tf.cast(image, tf.float32) * (1. / 255) - 0.5
label = tf.cast(features['label'], tf.int32)
return image, label

这段函数会返回一个图像和标签,我们需要按照Batch的方式读入

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
def inputs(train, batch_size, num_epochs):
if not num_epochs: num_epochs = None
filename = os.path.join(FLAGS.data_dir,
TRAIN_FILE if train else VALIDATION_FILE)

with tf.name_scope('input'):
filename_queue = tf.train.string_input_producer(
[filename], num_epochs=num_epochs)
image, label = read_and_decode(filename_queue)

images, sparse_labels = tf.train.shuffle_batch(
[image, label], batch_size=batch_size, num_threads=2,
capacity=1000 + 3 * batch_size,
min_after_dequeue=1000)

return images, sparse_labels

到这里我们可以读入batch图像和标签。

在CPU中定义变量

我们需要把weight和biases定义在CPU中,以便进行参数的更新。注意

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
```Python
def _variable_on_cpu(name, shape, initializer):
"""Helper to create a Variable stored on CPU memory.

Args:
name: name of the variable
shape: list of ints
initializer: initializer for Variable

Returns:
Variable Tensor
"""
with tf.device('/cpu:0'):
dtype = tf.float32
var = tf.get_variable(name, shape, initializer=initializer, dtype=dtype)
return var

构建Inference

构建Inference采用的的是卷积神经网络的架构,需要注意的是初始化的时候需要将变量定义在CPU中。

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
def inference(images):
"""Build the MNIST model.

Args:
images: Images returned from MNIST or inputs().

Returns:
Logits.
"""
x_image = tf.reshape(images, [-1, 28, 28, 1])

# conv1
with tf.variable_scope('conv1') as scope:

kernel = _variable_on_cpu('weights',shape=[5,5,1,32],
initializer = tf.truncated_normal_initializer(stddev=5e-2))
biases = _variable_on_cpu('biases', [32], tf.constant_initializer(0.0))
conv = tf.nn.conv2d(x_image, kernel, strides=[1, 1, 1, 1],
padding='SAME')
pre_activation = tf.nn.bias_add(conv, biases)
conv1 = tf.nn.relu(pre_activation, name=scope.name)


# pool1
pool1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
padding='SAME', name='pool1')

# conv2
with tf.variable_scope('conv2') as scope:

kernel = _variable_on_cpu('weights',shape=[5,5,32,64],
initializer = tf.truncated_normal_initializer(stddev=5e-2))
conv = tf.nn.conv2d(pool1, kernel, strides=[1, 1, 1, 1], padding='SAME')
biases = _variable_on_cpu('biases', [64], tf.constant_initializer(0.1))
pre_activation = tf.nn.bias_add(conv, biases)
conv2 = tf.nn.relu(pre_activation, name=scope.name)

# pool2
pool2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
padding='SAME', name='pool2')

# local3
with tf.variable_scope('local3') as scope:
# Move everything into depth so we can perform a single matrix multiply.
reshape = tf.reshape(pool2, [-1, 7 * 7 * 64])
dim = reshape.get_shape()[1].value

weights = _variable_on_cpu('weights',shape=[dim,1024],
initializer = tf.truncated_normal_initializer(stddev=0.04))
biases = _variable_on_cpu('biases', [1024],
tf.constant_initializer(0.1))
local3 = tf.nn.relu(tf.matmul(reshape, weights) + biases,
name=scope.name)

# local4
with tf.variable_scope('local4') as scope:
weights = _variable_on_cpu('weight',shape=[1024,10],
initializer = tf.truncated_normal_initializer(stddev=0.04))
biases = _variable_on_cpu('biases', [10], tf.constant_initializer(0.1))
local4 = tf.nn.relu(tf.matmul(local3, weights) + biases,
name=scope.name)

# linear layer(WX + b),
# We don't apply softmax here because
# tf.nn.sparse_softmax_cross_entropy_with_logits accepts the unscaled logits
# and performs the softmax internally for efficiency.
with tf.variable_scope('softmax_linear') as scope:

weights = _variable_on_cpu('weight',[10,10],
initializer = tf.truncated_normal_initializer(stddev=1 / 192.0))
biases = _variable_on_cpu('biases', [10],
tf.constant_initializer(0.0))
softmax_linear = tf.add(tf.matmul(local4, weights), biases,
name=scope.name)

return softmax_linear

定义Loss

定义loss的时候和单GPU的形式不同,因为我们不仅要定义损失函数,还要定义每个GPU的损失函数值和其梯度,最后再计算平均梯度。

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
def loss(logits, labels): 
"""Add L2Loss to all the trainable variables.

Add summary for "Loss" and "Loss/avg".
Args:
logits: Logits from inference().
labels: Labels from distorted_inputs or inputs(). 1-D tensor
of shape [batch_size]

Returns:
Loss tensor of type float.
"""
# Calculate the average cross entropy loss across the batch.
labels = tf.cast(labels, tf.int64)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
labels=labels, logits=logits, name='cross_entropy_per_example')
cross_entropy_mean = tf.reduce_mean(cross_entropy, name='cross_entropy')
tf.add_to_collection('losses', cross_entropy_mean)

# The total loss is defined as the cross entropy loss plus all of the weight
# decay terms (L2 loss).
return tf.add_n(tf.get_collection('losses'), name='total_loss')

def tower_loss(scope):
"""Calculate the total loss on a single tower running the MNIST model.

Args:
scope: unique prefix string identifying the MNIST tower, e.g. 'tower_0'

Returns:
Tensor of shape [] containing the total loss for a batch of data
"""
# Input images and labels.
images, labels = inputs(train=True, batch_size=FLAGS.batch_size,
num_epochs=FLAGS.num_epochs)
# Build inference Graph.
logits = inference(images)

# Build the portion of the Graph calculating the losses. Note that we will
# assemble the total_loss using a custom function below.
_ = loss(logits, labels)

# Assemble all of the losses for the current tower only.
losses = tf.get_collection('losses', scope)

# Calculate the total loss for the current tower.
total_loss = tf.add_n(losses, name='total_loss')

# Attach a scalar summary to all individual losses and the total loss; do
# the same for the averaged version of the losses.
if FLAGS.tb_logging:
for l in losses + [total_loss]:
# Remove 'tower_[0-9]/' from the name in case this is a multi-GPU
# training session. This helps the clarity of presentation on
# tensorboard.
loss_name = re.sub('%s_[0-9]*/' % TOWER_NAME, '', l.op.name)
tf.summary.scalar(loss_name, l)

return total_loss

def average_gradients(tower_grads):
"""Calculate average gradient for each shared variable across all towers.

Note that this function provides a synchronization point across all towers.

Args:
tower_grads: List of lists of (gradient, variable) tuples. The outer list
is over individual gradients. The inner list is over the gradient
calculation for each tower.
Returns:
List of pairs of (gradient, variable) where the gradient has been
averaged across all towers.
"""
average_grads = []
for grad_and_vars in zip(*tower_grads):
# Note that each grad_and_vars looks like the following:
# ((grad0_gpu0, var0_gpu0), ... , (grad0_gpuN, var0_gpuN))
grads = []
for g, _ in grad_and_vars:
# Add 0 dimension to the gradients to represent the tower.
expanded_g = tf.expand_dims(g, 0)

# Append on a 'tower' dimension which we will average over below.
grads.append(expanded_g)

# Average over the 'tower' dimension.
grad = tf.concat(grads, 0)
grad = tf.reduce_mean(grad, 0)

# Keep in mind that the Variables are redundant because they are shared
# across towers. So .. we will just return the first tower's pointer to
# the Variable.
v = grad_and_vars[0][1]
grad_and_var = (grad, v)
average_grads.append(grad_and_var)
return average_grads

定义训练过程

训练过程的需要注意把不同的环节放在不同的devices下面。

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
def train():
with tf.Graph().as_default(), tf.device('/cpu:0'):
# Create a variable to count the number of train() calls. This equals
# the number of batches processed * FLAGS.num_gpus.
global_step = tf.get_variable(
'global_step', [],
initializer=tf.constant_initializer(0), trainable=False)

# opt = tf.train.MomentumOptimizer(lr,0.9,use_nesterov=True,use_locking=True)
opt = tf.train.MomentumOptimizer(INITIAL_LEARNING_RATE,0.9,use_nesterov=True,use_locking=True)

# Calculate the gradients for each model tower.
tower_grads = []
with tf.variable_scope(tf.get_variable_scope()):
for i in xrange(FLAGS.num_gpus):
with tf.device('/gpu:%d' % i):
with tf.name_scope(
'%s_%d' % (TOWER_NAME, i)) as scope:
# Calculate the loss for one tower of the CIFAR model.
# This function constructs the entire CIFAR model but
# shares the variables across all towers.
loss = tower_loss(scope)

# Reuse variables for the next tower.
tf.get_variable_scope().reuse_variables()

# Retain the summaries from the final tower.
summaries = tf.get_collection(tf.GraphKeys.SUMMARIES,
scope)

# Calculate the gradients for the batch of data on this
# MNIST tower.
grads = opt.compute_gradients(loss, gate_gradients=0)

# Keep track of the gradients across all towers.
tower_grads.append(grads)

# We must calculate the mean of each gradient. Note that this is the
# synchronization point across all towers.
grads = average_gradients(tower_grads)

train_op = opt.apply_gradients(grads, global_step=global_step)


# The op for initializing the variables.
init_op = tf.group(tf.global_variables_initializer(),
tf.local_variables_initializer())

# Start running operations on the Graph. allow_soft_placement must be
# set to True to build towers on GPU, as some of the ops do not have GPU
# implementations.
sess = tf.Session(config=tf.ConfigProto(
allow_soft_placement=True,
log_device_placement=FLAGS.log_device_placement))
sess.run(init_op)

# Start input enqueue threads.
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)

try:
step = 0
while not coord.should_stop():
start_time = time.time()

# Run one step of the model. The return values are
# the activations from the `train_op` (which is
# discarded) and the `loss` op. To inspect the values
# of your ops or variables, you may include them in
# the list passed to sess.run() and the value tensors
# will be returned in the tuple from the call.
_, loss_value = sess.run([train_op, loss])

duration = time.time() - start_time

# assert not np.isnan(
# loss_value), 'Model diverged with loss = NaN'

# Print an overview fairly often.
if step % 100 == 0:
num_examples_per_step = FLAGS.batch_size * FLAGS.num_gpus
examples_per_sec = num_examples_per_step / duration
sec_per_batch = duration / FLAGS.num_gpus
format_str = (
'%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print(format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
step += 1
except tf.errors.OutOfRangeError:
print('Done training for %d epochs, %d steps.' % (
FLAGS.num_epochs, step))
finally:
# When done, ask the threads to stop.
coord.request_stop()

# Wait for threads to finish.
coord.join(threads)
sess.close()

最后就可以调用Train()函数进行训练了。训练函数分配GPU的时候有for循环,所以可以支持不同数量的GPU。

单机多卡服务器进行深度学习的训练,构建代码比较复杂,并且需要手动分配devices,相比于NCCL的高级库好的一点就是可以针对不同的任务进行定制化的分配,以实现最大程度的优化,工作量比较大,效果也非常好。搭建的时候需要平衡一下效率和开发速度。后续还会尝试多机多卡的情况,目前还在尝试。