技术标签: tensorflow 源码 深度学习与计算机视觉
标签(空格分隔): tensorflow 源码
caffe如果用到多GPU训练,则需要进行编译配置MPI,而且很容易报错,tensorflow只需要在代码中使用即可。
官方给的代码:
# 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_function
from datetime import datetime
import os.path
import re
import time
import numpy as np
from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf
import cifar10
FLAGS = tf.app.flags.FLAGS
tf.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.""")
#使用多少个GPU
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)# 建立网络 并输出分类结构
# 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) #计算损失函数
# Assemble all of the losses for the current tower only.
losses = tf.get_collection('losses', scope) #获取损失值
# Calculate the total loss for the current tower.
total_loss = tf.add_n(losses, name='total_loss') # 获取总共的损失值,可能还有正则项啥的
# Attach a scalar summary to all individual losses and the total loss; do the
# same for the averaged version of the losses.
#然后 进行统计记录各个损失值
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_loss
def 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 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(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_grads
def train():
"""Train CIFAR-10 for a number of steps."""
with tf.Graph().as_default(), tf.device('/cpu:0'):#建立图模型
# 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.
#设置学习率的衰减方式,优化函数,数据读入等等
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 = [] # 计算每个GPU中 图模型的梯度存放在这个list中
with tf.variable_scope(tf.get_variable_scope()):
for i in xrange(FLAGS.num_gpus):#循环建立 网络图模型,设置所用的GPU的编号
with tf.device('/gpu:%d' % i):
with tf.name_scope('%s_%d' % (cifar10.TOWER_NAME, i)) as scope:
# Dequeues one batch for the GPU
image_batch, label_batch = batch_queue.dequeue()#读取数据
# 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)# 在该GPU中建立网络,计算损失函数
# Reuse variables for the next tower.
tf.get_variable_scope().reuse_variables()#下一次就重复使用该图的权值,即reuse=True
# 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)#计算当前GPU下各个训练参数的梯度
# Keep track of the gradients across all towers.
tower_grads.append(grads)# 然后加入到这个list中
# We must calculate the mean of each gradient. Note that this is the
# synchronization point across all towers.
grads = average_gradients(tower_grads)#在所用GPUs下,求取平均梯度
# 进行一系列的统计,记录,到tensorboard中
# 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.
#将梯度应用到各个训练参数中
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.
#权值平移,平均化(不知道啥意思,反正就是保证梯度的)
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.
# 优化op
train_op = tf.group(apply_gradient_op, variables_averages_op)
# Create a saver.
#存储变量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)
# 后面就是各种seess.run,训练的显示,测试啥的
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 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))
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-argument
cifar10.maybe_download_and_extract()
if tf.gfile.Exists(FLAGS.train_dir):
tf.gfile.DeleteRecursively(FLAGS.train_dir)
tf.gfile.MakeDirs(FLAGS.train_dir)
train()#核心函数
if __name__ == '__main__':
tf.app.run()
只是官方所给的多GPU训练demo,直接把数据cifar10下载下来,就可以美滋滋的训练了!
稍后会给出笔者采用多GPU的另外方式
文章浏览阅读1k次。通过使用ajax方法跨域请求是浏览器所不允许的,浏览器出于安全考虑是禁止的。警告信息如下:不过jQuery对跨域问题也有解决方案,使用jsonp的方式解决,方法如下:$.ajax({ async:false, url: 'http://www.mysite.com/demo.do', // 跨域URL ty..._nginx不停的xhr
文章浏览阅读2k次。关于在 Oracle 中配置 extproc 以访问 ST_Geometry,也就是我们所说的 使用空间SQL 的方法,官方文档链接如下。http://desktop.arcgis.com/zh-cn/arcmap/latest/manage-data/gdbs-in-oracle/configure-oracle-extproc.htm其实简单总结一下,主要就分为以下几个步骤。..._extproc
文章浏览阅读1.5w次。linux下没有上面的两个函数,需要使用函数 mbstowcs和wcstombsmbstowcs将多字节编码转换为宽字节编码wcstombs将宽字节编码转换为多字节编码这两个函数,转换过程中受到系统编码类型的影响,需要通过设置来设定转换前和转换后的编码类型。通过函数setlocale进行系统编码的设置。linux下输入命名locale -a查看系统支持的编码_linux c++ gbk->utf8
文章浏览阅读750次。今天准备从生产库向测试库进行数据导入,结果在imp导入的时候遇到“ IMP-00009:导出文件异常结束” 错误,google一下,发现可能有如下原因导致imp的数据太大,没有写buffer和commit两个数据库字符集不同从低版本exp的dmp文件,向高版本imp导出的dmp文件出错传输dmp文件时,文件损坏解决办法:imp时指定..._imp-00009导出文件异常结束
文章浏览阅读143次。当下是一个大数据的时代,各个行业都离不开数据的支持。因此,网络爬虫就应运而生。网络爬虫当下最为火热的是Python,Python开发爬虫相对简单,而且功能库相当完善,力压众多开发语言。本次教程我们爬取前程无忧的招聘信息来分析Python程序员需要掌握那些编程技术。首先在谷歌浏览器打开前程无忧的首页,按F12打开浏览器的开发者工具。浏览器开发者工具是用于捕捉网站的请求信息,通过分析请求信息可以了解请..._初级python程序员能力要求
文章浏览阅读7.6k次,点赞2次,收藏6次。@Service标注的bean,类名:ABDemoService查看源码后发现,原来是经过一个特殊处理:当类的名字是以两个或以上的大写字母开头的话,bean的名字会与类名保持一致public class AnnotationBeanNameGenerator implements BeanNameGenerator { private static final String C..._@service beanname
文章浏览阅读6.9w次,点赞73次,收藏463次。1.前序创建#include<stdio.h>#include<string.h>#include<stdlib.h>#include<malloc.h>#include<iostream>#include<stack>#include<queue>using namespace std;typed_二叉树的建立
文章浏览阅读7.1k次。在Asp.net上使用Excel导出功能,如果文件名出现中文,便会以乱码视之。 解决方法: fileName = HttpUtility.UrlEncode(fileName, System.Text.Encoding.UTF8);_asp.net utf8 导出中文字符乱码
文章浏览阅读2.1k次,点赞4次,收藏23次。第一次实验 词法分析实验报告设计思想词法分析的主要任务是根据文法的词汇表以及对应约定的编码进行一定的识别,找出文件中所有的合法的单词,并给出一定的信息作为最后的结果,用于后续语法分析程序的使用;本实验针对 PL/0 语言 的文法、词汇表编写一个词法分析程序,对于每个单词根据词汇表输出: (单词种类, 单词的值) 二元对。词汇表:种别编码单词符号助记符0beginb..._对pl/0作以下修改扩充。增加单词
文章浏览阅读773次。我在使用adb.exe时遇到了麻烦.我想使用与bash相同的adb.exe shell提示符,所以我决定更改默认的bash二进制文件(当然二进制文件是交叉编译的,一切都很完美)更改bash二进制文件遵循以下顺序> adb remount> adb push bash / system / bin /> adb shell> cd / system / bin> chm..._adb shell mv 权限
文章浏览阅读6.8k次,点赞12次,收藏125次。1. 单目相机标定引言相机标定已经研究多年,标定的算法可以分为基于摄影测量的标定和自标定。其中,应用最为广泛的还是张正友标定法。这是一种简单灵活、高鲁棒性、低成本的相机标定算法。仅需要一台相机和一块平面标定板构建相机标定系统,在标定过程中,相机拍摄多个角度下(至少两个角度,推荐10~20个角度)的标定板图像(相机和标定板都可以移动),即可对相机的内外参数进行标定。下面介绍张氏标定法(以下也这么称呼)的原理。原理相机模型和单应矩阵相机标定,就是对相机的内外参数进行计算的过程,从而得到物体到图像的投影_相机-投影仪标定
文章浏览阅读2.2k次。文章目录Wayland 架构Wayland 渲染Wayland的 硬件支持简 述: 翻译一篇关于和 wayland 有关的技术文章, 其英文标题为Wayland Architecture .Wayland 架构若是想要更好的理解 Wayland 架构及其与 X (X11 or X Window System) 结构;一种很好的方法是将事件从输入设备就开始跟踪, 查看期间所有的屏幕上出现的变化。这就是我们现在对 X 的理解。 内核是从一个输入设备中获取一个事件,并通过 evdev 输入_wayland