深度学习theano/tensorflow多显卡多人使用问题集

tensorflow中使用指定的GPU及GPU显存

Using GPUs

petewarden/tensorflow_makefile

tf_gpu_manager/manager.py

多GPU运行Deep Learning 和并行Deep Learning（待续）

Multiple GPUs

这里写图片描述

1. 终端执行程序时设置使用的GPU

如果电脑有多个GPU，tensorflow默认全部使用。如果想只使用部分GPU，可以设置CUDA_VISIBLE_DEVICES。在调用python程序时，可以使用

CUDA_VISIBLE_DEVICES=1 python my_script.py #只使用GPU1
CUDA_VISIBLE_DEVICES=0,1 python my_script.py #使用GPU0,GPU1

Environment Variable Syntax      ResultsCUDA_VISIBLE_DEVICES=1           Only device 1 will be seen
CUDA_VISIBLE_DEVICES=0,1         Devices 0 and 1 will be visible
CUDA_VISIBLE_DEVICES="0,1"       Same as above, quotation marks are optional
CUDA_VISIBLE_DEVICES=0,2,3       Devices 0, 2, 3 will be visible; device 1 is masked
CUDA_VISIBLE_DEVICES=""          No GPU will be visible

2. python代码中设置使用的GPU

如果要在python代码中设置使用的GPU，可以使用下面的代码

import os
os.environ["CUDA_VISIBLE_DEVICES"] = "2"

3. 设置tensorflow使用的显存大小

定量设置显存

默认tensorflow是使用GPU尽可能多的显存。可以通过下面的方式，来设置使用的GPU显存：

gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.7)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))

上面分配给tensorflow的GPU显存大小为：GPU实际显存*0.7。
可以按照需要，设置不同的值，来分配显存。

按需设置显存

上面的只能设置固定的大小。如果想按需分配，可以使用allow_growth参数（参考网址：http://blog.csdn.net/cq361106306/article/details/52950081）：

gpu_options = tf.GPUOptions(allow_growth=True)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))

4. 使用多个 GPU

如果你想让 TensorFlow 在多个 GPU 上运行, 你可以建立 multi-tower 结构, 在这个结构里每个 tower 分别被指配给不同的 GPU 运行. 比如:

# 新建一个 graph.
c = []
for d in ['/gpu:2', '/gpu:3']:with tf.device(d):a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3])b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2])c.append(tf.matmul(a, b))
with tf.device('/cpu:0'):sum = tf.add_n(c)
# 新建session with log_device_placement并设置为True.
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
# 运行这个op.
print sess.run(sum)

你会看到如下输出:

Device mapping:
/job:localhost/replica:0/task:0/gpu:0 -> device: 0, name: Tesla K20m, pci bus
id: 0000:02:00.0
/job:localhost/replica:0/task:0/gpu:1 -> device: 1, name: Tesla K20m, pci bus
id: 0000:03:00.0
/job:localhost/replica:0/task:0/gpu:2 -> device: 2, name: Tesla K20m, pci bus
id: 0000:83:00.0
/job:localhost/replica:0/task:0/gpu:3 -> device: 3, name: Tesla K20m, pci bus
id: 0000:84:00.0
Const_3: /job:localhost/replica:0/task:0/gpu:3
Const_2: /job:localhost/replica:0/task:0/gpu:3
MatMul_1: /job:localhost/replica:0/task:0/gpu:3
Const_1: /job:localhost/replica:0/task:0/gpu:2
Const: /job:localhost/replica:0/task:0/gpu:2
MatMul: /job:localhost/replica:0/task:0/gpu:2
AddN: /job:localhost/replica:0/task:0/cpu:0
[[  44.   56.][  98.  128.]]

5. 如何实现multi_gpu_model函数

def multi_gpu_model(num_gpus=1):grads = []for i in range(num_gpus):with tf.device("/gpu:%d"%i):with tf.name_scope("tower_%d"%i):model = Model(is_training, config, scope)# 放到collection中，方便feed的时候取tf.add_to_collection("train_model", model)grads.append(model.grad) #grad 是通过tf.gradients(loss, vars)求得#以下这些add_to_collection可以直接在模型内部完成。# 将loss放到 collection中， 方便以后操作tf.add_to_collection("loss",model.loss)#将predict放到collection中，方便操作tf.add_to_collection("predict", model.predict)#将 summary.merge op放到collection中，方便操作tf.add_to_collection("merge_summary", model.merge_summary)# ...with tf.device("cpu:0"):averaged_gradients = average_gradients(grads)# average_gradients后面说明opt = tf.train.GradientDescentOptimizer(learning_rate)train_op=opt.apply_gradients(zip(average_gradients,tf.trainable_variables()))return train_op

6. cifar10 tutorial-cifar10_multi_gpu_train.py

code 见 models/tutorials/image/cifar10/

# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# =============================================================================="""A binary to train CIFAR-10 using multiple GPUs with synchronous updates.
Accuracy:
cifar10_multi_gpu_train.py achieves ~86% accuracy after 100K steps (256
epochs of data) as judged by cifar10_eval.py.
Speed: With batch_size 128.
System        | Step Time (sec/batch)  |     Accuracy
--------------------------------------------------------------------
1 Tesla K20m  | 0.35-0.60              | ~86% at 60K steps  (5 hours)
1 Tesla K40m  | 0.25-0.35              | ~86% at 100K steps (4 hours)
2 Tesla K20m  | 0.13-0.20              | ~84% at 30K steps  (2.5 hours)
3 Tesla K20m  | 0.13-0.18              | ~84% at 30K steps
4 Tesla K20m  | ~0.10                  | ~84% at 30K steps
Usage:
Please see the tutorial and website for how to download the CIFAR-10
data set, compile the program and train the model.
http://tensorflow.org/tutorials/deep_cnn/
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_functionfrom datetime import datetime
import os.path
import re
import timeimport numpy as np
from six.moves import xrange  # pylint: disable=redefined-builtin
import tensorflow as tf
import cifar10FLAGS = tf.app.flags.FLAGStf.app.flags.DEFINE_string('train_dir', '/tmp/cifar10_train',"""Directory where to write event logs """"""and checkpoint.""")
tf.app.flags.DEFINE_integer('max_steps', 1000000,"""Number of batches to run.""")
tf.app.flags.DEFINE_integer('num_gpus', 1,"""How many GPUs to use.""")
tf.app.flags.DEFINE_boolean('log_device_placement', False,"""Whether to log device placement.""")def tower_loss(scope, images, labels):"""Calculate the total loss on a single tower running the CIFAR model.Args:scope: unique prefix string identifying the CIFAR tower, e.g. 'tower_0'images: Images. 4D tensor of shape [batch_size, height, width, 3].labels: Labels. 1D tensor of shape [batch_size].Returns:Tensor of shape [] containing the total loss for a batch of data"""# Build inference Graph.logits = cifar10.inference(images)# Build the portion of the Graph calculating the losses. Note that we will# assemble the total_loss using a custom function below._ = cifar10.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.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]*/' % cifar10.TOWER_NAME, '', l.op.name)tf.summary.scalar(loss_name, l)return total_lossdef average_gradients(tower_grads):"""Calculate the 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 listis over individual gradients. The inner list is over the gradientcalculation for each tower.Returns:List of pairs of (gradient, variable) where the gradient has been averagedacross 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(axis=0, values=grads)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_gradsdef train():"""Train CIFAR-10 for a number of steps."""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)# Calculate the learning rate schedule.num_batches_per_epoch = (cifar10.NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN /FLAGS.batch_size)decay_steps = int(num_batches_per_epoch * cifar10.NUM_EPOCHS_PER_DECAY)# Decay the learning rate exponentially based on the number of steps.lr = tf.train.exponential_decay(cifar10.INITIAL_LEARNING_RATE,global_step,decay_steps,cifar10.LEARNING_RATE_DECAY_FACTOR,staircase=True)# Create an optimizer that performs gradient descent.opt = tf.train.GradientDescentOptimizer(lr)# Get images and labels for CIFAR-10.images, labels = cifar10.distorted_inputs()batch_queue = tf.contrib.slim.prefetch_queue.prefetch_queue([images, labels], capacity=2 * FLAGS.num_gpus)# 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' % (cifar10.TOWER_NAME, i)) as scope:# Dequeues one batch for the GPUimage_batch, label_batch = batch_queue.dequeue()# 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, image_batch, label_batch)# 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 CIFAR tower.grads = opt.compute_gradients(loss)# 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)# Add a summary to track the learning rate.summaries.append(tf.summary.scalar('learning_rate', lr))# Add histograms for gradients.for grad, var in grads:if grad is not None:summaries.append(tf.summary.histogram(var.op.name + '/gradients', grad))# Apply the gradients to adjust the shared variables.apply_gradient_op = opt.apply_gradients(grads, global_step=global_step)# Add histograms for trainable variables.for var in tf.trainable_variables():summaries.append(tf.summary.histogram(var.op.name, var))# Track the moving averages of all trainable variables.variable_averages = tf.train.ExponentialMovingAverage(cifar10.MOVING_AVERAGE_DECAY, global_step)variables_averages_op = variable_averages.apply(tf.trainable_variables())# Group all updates to into a single train op.train_op = tf.group(apply_gradient_op, variables_averages_op)# Create a saver.saver = tf.train.Saver(tf.global_variables())# Build the summary operation from the last tower summaries.summary_op = tf.summary.merge(summaries)# Build an initialization operation to run below.init = tf.global_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)# Start the queue runners.tf.train.start_queue_runners(sess=sess)summary_writer = tf.summary.FileWriter(FLAGS.train_dir, sess.graph)for step in xrange(FLAGS.max_steps):start_time = time.time()_, loss_value = sess.run([train_op, loss])duration = time.time() - start_timeassert not np.isnan(loss_value), 'Model diverged with loss = NaN'if step % 10 == 0:num_examples_per_step = FLAGS.batch_size * FLAGS.num_gpusexamples_per_sec = num_examples_per_step / durationsec_per_batch = duration / FLAGS.num_gpusformat_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))if step % 100 == 0:summary_str = sess.run(summary_op)summary_writer.add_summary(summary_str, step)# Save the model checkpoint periodically.if step % 1000 == 0 or (step + 1) == FLAGS.max_steps:checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt')saver.save(sess, checkpoint_path, global_step=step)def main(argv=None):  # pylint: disable=unused-argumentcifar10.maybe_download_and_extract()if tf.gfile.Exists(FLAGS.train_dir):tf.gfile.DeleteRecursively(FLAGS.train_dir)tf.gfile.MakeDirs(FLAGS.train_dir)train()if __name__ == '__main__':
tf.app.run()

python cifar10_multi_gpu_train.py --num_gpus=2

参考文献

http://stackoverflow.com/questions/36668467/change-default-gpu-in-tensorflow
http://stackoverflow.com/questions/37893755/tensorflow-set-cuda-visible-devices-within-jupyter
（原）tensorflow中使用指定的GPU及GPU显存
Using GPUs
TensorFlow官方文档中文版 » 运作方式 » 使用gpu
tensorflow学习笔记（三十一）：构建多GPU代码
cifar10 tutorial
CIFAR10 多 GPU 版本例程源码分析
tensorflow cifar_10 代码阅读与理解