Caffe学习:Solver

来源:互联网 发布:苹果糯米网络不可用 编辑:程序博客网 时间:2024/06/06 01:53

目录:

  • 原文
  • Solver
  • Methods
    • SGD
    • AdaGrad
    • NAG
  • Scaffolding
  • Updating Parameters
  • Snapshotting and Resuming

原文


Solver

Solver通过协调Net的前向loss计算和反向梯度计算(forward inference and backward gradients),对参数进行更新,以减小loss。Caffe模型的学习被分为两个部分:由Solver进行优化、更新参数,由Net计算出loss和gradient。

Caffe提供了3种Solver方法:Stochastic Gradient Descent(SGD,随机梯度下降),Adaptive Gradient(ADAGRAD,自适应梯度下降)和Nesterov’s Accelerated Gradient(NESTEROV,Nesterov提出的加速梯度下降)。

Solve步骤:

  1. 设计优化对象、训练网络(用于学习)和测试网络(用于评估)
  2. 通过forward和backward迭代以优化和更新参数
  3. 周期性地评估测试网络
  4. 在优化过程中显示(snapshot)模型和solver的状态

每一次迭代中:

  1. 调用Net的forward pass计算出output 和loss
  2. 调用Net的backward pass计算出gradients
  3. 根据Solver方法,利用gradients更新参数
  4. 根据learning rate,history和method更新solver的状态

Caffe solvers有CPU / GPU两种模式。


Methods

Solver Methods用于最小化loss值。给定一个数据集D,优化的目标是D中所有数据的平均loss,即使平均loss取最小值。

计算公式

其中 fW(X(i))是data中X(i)项的loss,r(W)是正则项,权重为λ。当数据量很大时(直接采用普通的梯度下降方法计算量很大),在每一次迭代中,我们采用数据集的一个子集(mini-batch)来近似代替,其数据量远小于整个数据集:

计算公式

在forward pass中计算fW(即loss),在backward pass中计算∇fW(即gradient)。

SGD

  • 类型:SGD

随机梯度下降(Stochastic gradient)利用(负的)gradient ∇L(W) 上一次权重的更新值Vt的线性组合来更新权重W学习速率(learning rate)α是(负的)gradient的权重。动量(momentum )μ是上一次更新值得权重。

有如下公式,根据上一次计算的更新值Vt和当前权重来计算本次的更新值Vt+1和权重 Wt+1:

计算公式

学习的参数(α和μ)需要一定的调整才能达到最好的效果。如果你不是很清楚该怎样做的话,请看下面的“经验”,更多的细节请参考Leon Bottou的Stochastic Gradient Descent Tricks

设置学习速率( learning rate α)和动量(momentum μ)的经验:

一个比较好的建议是,将学习速率( learning rate α)初始化为α≈0.01,然后在训练(training)中当loss达到稳定时,将α除以一个常数(例如,10)。对于,动量(momentum μ)一般设置为μ=0.9,μ使weight的更新更为平缓,是学习过程更为稳定、快速。

这是Krizhevsky提出的技巧【文献1】,Caffe实现了SolverParameter让这个技巧易于实现。详见:./examples/imagenet/alexnet_solver.prototxt

要使用上述技巧,可以将下面的代码添加到自定义的solver prototxt文件中:

<code class="hljs avrasm has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;"><span class="hljs-label" style="box-sizing: border-box;">base_lr:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.01</span>     <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 开始学习速率为:α = 0.01</span><span class="hljs-label" style="box-sizing: border-box;">lr_policy:</span> <span class="hljs-string" style="color: rgb(0, 136, 0); box-sizing: border-box;">"step"</span> <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 学习策略: 每stepsize次迭代之后,将α乘以gamma</span><span class="hljs-label" style="box-sizing: border-box;">gamma:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.1</span>        <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 学习速率变化因子</span><span class="hljs-label" style="box-sizing: border-box;">stepsize:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">100000</span>  <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 每100K次迭代,下降学习速率</span><span class="hljs-label" style="box-sizing: border-box;">max_iter:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">350000</span>  <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 训练的最大迭代次数</span><span class="hljs-label" style="box-sizing: border-box;">momentum:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.9</span>     <span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># momentum为:μ = 0.01</span></code><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li></ul><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li></ul>

上面的例子中,我们将动量μ设为常数0.9。刚开始前100K次迭代时学习速率α设置为0.01,然后第100K~200K次迭代时将学习速率α乘以gamma (γ),即为α′=αγ=(0.01)*(0.1)=0.001,之后第200K~300K次迭代时学习速率为α′′=0.00001,最后第301K~350K次迭代时学习速率为α′′=0.000001.

上述例子中,当训练次数达到一定量后,计算量会扩大到1/(1-μ)倍,所以如果增加μ的话,最好是减少α值(反之相反)。举个例子, 设μ=0.9,则计算量会扩大1/(1-μ)=10倍。如果将μ扩大为0.99,那么计算量会扩大100倍,所以α应该除以10.

上述技巧也只是经验之谈,不保证绝对有用,甚至有副作用。如果训练过程中出现了发散现象(例如,loss或者output值非常大),试着减小开始学习速率(base_lr)再训练,重复这个过程,直到找到一个比较合适的学习速率。

【文献1】:A. Krizhevsky, I. Sutskever, and G. Hinton. ImageNet Classification with Deep Convolutional Neural Networks. Advances in Neural Information Processing Systems, 2012.

AdaGrad

  • 类型:ADAGRAD

自适应梯度下降方法( Adaptive gradient)【文献1】跟随机梯度下降(Stochastic gradient)一样是基于梯度的优化方法。给定之前更新的信息:之前的更新信息,按照下面的公式更新参数: 
计算公式

实际操作中,对于权重权重,自适应梯度(AdaGrad )只需要存储空间额外空间存储历史信息,而不是存储空间的存储空间来存储所有的历史纪录。

【文献1】: Duchi, E. Hazan, and Y. Singer. Adaptive Subgradient Methods for Online Learning and Stochastic Optimization. The Journal of Machine Learning Research, 2011.

NAG

  • 类型:NESTEROV

Nesterov提出的加速梯度下降(Nesterov’s accelerated gradient)是Nesterov提出的凸优化的最佳方法【文献1】,其收敛速度可以达到收敛速度,而不是收敛速度。尽管在Caffe中并不是一定能够满足收敛于收敛速度的条件(例如,由于非平滑non-smoothness和non-convexity非凸)实际中NAG对于某类特定结构的深度学习模型是一个非常有效的方法【文献2】。

权重weight更新参数与随机梯度下降(Stochastic gradient)非常相似:

计算公式

与SGD的不同之处在于梯度∇L(W)项中取值不同:在NAG中,对于权重W梯度的计算要机上动量=>∇L(Wt+μVt);在SGD中,只是简单的计算当前权重的动量=>∇L(Wt)。

【文献1】:Y. Nesterov. A Method of Solving a Convex Programming Problem with Convergence Rate O(1/k√). Soviet Mathematics Doklady, 1983. 
【文献2】:I. Sutskever, J. Martens, G. Dahl, and G. Hinton. On the Importance of Initialization and Momentum in Deep Learning. Proceedings of the 30th International Conference on Machine Learning, 2013.


Scaffolding

Solver::Presolve()方法:准备用于优化的方法及初始化模型。

<code class="hljs avrasm has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;">> caffe train -solver examples/mnist/lenet_solver<span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;">.prototxt</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56.474978</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> caffe<span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;">.cpp</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">90</span>] Starting OptimizationI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56.475190</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> solver<span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;">.cpp</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">32</span>] Initializing solver from parameters:<span class="hljs-label" style="box-sizing: border-box;">test_iter:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">100</span><span class="hljs-label" style="box-sizing: border-box;">test_interval:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">500</span><span class="hljs-label" style="box-sizing: border-box;">base_lr:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.01</span><span class="hljs-label" style="box-sizing: border-box;">display:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">100</span><span class="hljs-label" style="box-sizing: border-box;">max_iter:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">10000</span><span class="hljs-label" style="box-sizing: border-box;">lr_policy:</span> <span class="hljs-string" style="color: rgb(0, 136, 0); box-sizing: border-box;">"inv"</span><span class="hljs-label" style="box-sizing: border-box;">gamma:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.0001</span><span class="hljs-label" style="box-sizing: border-box;">power:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.75</span><span class="hljs-label" style="box-sizing: border-box;">momentum:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.9</span><span class="hljs-label" style="box-sizing: border-box;">weight_decay:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0.0005</span><span class="hljs-label" style="box-sizing: border-box;">snapshot:</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5000</span><span class="hljs-label" style="box-sizing: border-box;">snapshot_prefix:</span> <span class="hljs-string" style="color: rgb(0, 136, 0); box-sizing: border-box;">"examples/mnist/lenet"</span><span class="hljs-label" style="box-sizing: border-box;">solver_mode:</span> GPU<span class="hljs-label" style="box-sizing: border-box;">net:</span> <span class="hljs-string" style="color: rgb(0, 136, 0); box-sizing: border-box;">"examples/mnist/lenet_train_test.prototxt"</span></code><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li><li style="box-sizing: border-box; padding: 0px 5px;">15</li><li style="box-sizing: border-box; padding: 0px 5px;">16</li><li style="box-sizing: border-box; padding: 0px 5px;">17</li></ul><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li><li style="box-sizing: border-box; padding: 0px 5px;">15</li><li style="box-sizing: border-box; padding: 0px 5px;">16</li><li style="box-sizing: border-box; padding: 0px 5px;">17</li></ul>

Net initialization(网络初始化)

<code class="hljs dos has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;">I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">655681</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> solver.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">72</span>] Creating training <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span> from <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span> file: examples/mnist/lenet_train_test.prototxt[...]I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">656740</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">656791</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">67</span>] Creating Layer mnistI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">656811</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">356</span>] mnist -> dataI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">656846</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">356</span>] mnist -> labelI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">656874</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">96</span>] Setting up mnistI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">694052</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> data_layer.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">135</span>] Opening lmdb examples/mnist/mnist_train_lmdbI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">701062</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> data_layer.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">195</span>] output data size: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span>,<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span>,<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">28</span>,<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">28</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">701146</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> data_layer.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">236</span>] Initializing prefetchI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">701196</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> data_layer.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">238</span>] Prefetch initialized.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">701212</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">28</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">28</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">50176</span>)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">701230</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span>)[...]I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">703737</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">67</span>] Creating Layer ip1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">703753</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">394</span>] ip1 <- pool2I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">703778</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">356</span>] ip1 -> ip1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">703797</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">96</span>] Setting up ip1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728127</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">500</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">32000</span>)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728142</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">113</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5039360</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728175</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">67</span>] Creating Layer relu1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728194</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">394</span>] relu1 <- ip1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728219</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">345</span>] relu1 -> ip1 (<span class="hljs-flow" style="box-sizing: border-box;">in</span>-place)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728240</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">96</span>] Setting up relu1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728256</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">500</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">32000</span>)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728270</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">113</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5167360</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728287</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">67</span>] Creating Layer ip2I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728304</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">394</span>] ip2 <- ip1I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728333</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">356</span>] ip2 -> ip2I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728356</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">96</span>] Setting up ip2I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728690</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">64</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">10</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">640</span>)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728705</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">113</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5169920</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728734</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">67</span>] Creating Layer lossI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728747</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">394</span>] loss <- ip2I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728767</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">394</span>] loss <- labelI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728786</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">356</span>] loss -> lossI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728811</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">96</span>] Setting up lossI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728837</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">103</span>] Top shape: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span> (<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span>)I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728849</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">109</span>]     with loss weight <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">1</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728878</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">113</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5169924</span></code><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li><li style="box-sizing: border-box; padding: 0px 5px;">15</li><li style="box-sizing: border-box; padding: 0px 5px;">16</li><li style="box-sizing: border-box; padding: 0px 5px;">17</li><li style="box-sizing: border-box; padding: 0px 5px;">18</li><li style="box-sizing: border-box; padding: 0px 5px;">19</li><li style="box-sizing: border-box; padding: 0px 5px;">20</li><li style="box-sizing: border-box; padding: 0px 5px;">21</li><li style="box-sizing: border-box; padding: 0px 5px;">22</li><li style="box-sizing: border-box; padding: 0px 5px;">23</li><li style="box-sizing: border-box; padding: 0px 5px;">24</li><li style="box-sizing: border-box; padding: 0px 5px;">25</li><li style="box-sizing: border-box; padding: 0px 5px;">26</li><li style="box-sizing: border-box; padding: 0px 5px;">27</li><li style="box-sizing: border-box; padding: 0px 5px;">28</li><li style="box-sizing: border-box; padding: 0px 5px;">29</li><li style="box-sizing: border-box; padding: 0px 5px;">30</li><li style="box-sizing: border-box; padding: 0px 5px;">31</li><li style="box-sizing: border-box; padding: 0px 5px;">32</li><li style="box-sizing: border-box; padding: 0px 5px;">33</li><li style="box-sizing: border-box; padding: 0px 5px;">34</li><li style="box-sizing: border-box; padding: 0px 5px;">35</li><li style="box-sizing: border-box; padding: 0px 5px;">36</li><li style="box-sizing: border-box; padding: 0px 5px;">37</li><li style="box-sizing: border-box; padding: 0px 5px;">38</li><li style="box-sizing: border-box; padding: 0px 5px;">39</li><li style="box-sizing: border-box; padding: 0px 5px;">40</li></ul><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li><li style="box-sizing: border-box; padding: 0px 5px;">15</li><li style="box-sizing: border-box; padding: 0px 5px;">16</li><li style="box-sizing: border-box; padding: 0px 5px;">17</li><li style="box-sizing: border-box; padding: 0px 5px;">18</li><li style="box-sizing: border-box; padding: 0px 5px;">19</li><li style="box-sizing: border-box; padding: 0px 5px;">20</li><li style="box-sizing: border-box; padding: 0px 5px;">21</li><li style="box-sizing: border-box; padding: 0px 5px;">22</li><li style="box-sizing: border-box; padding: 0px 5px;">23</li><li style="box-sizing: border-box; padding: 0px 5px;">24</li><li style="box-sizing: border-box; padding: 0px 5px;">25</li><li style="box-sizing: border-box; padding: 0px 5px;">26</li><li style="box-sizing: border-box; padding: 0px 5px;">27</li><li style="box-sizing: border-box; padding: 0px 5px;">28</li><li style="box-sizing: border-box; padding: 0px 5px;">29</li><li style="box-sizing: border-box; padding: 0px 5px;">30</li><li style="box-sizing: border-box; padding: 0px 5px;">31</li><li style="box-sizing: border-box; padding: 0px 5px;">32</li><li style="box-sizing: border-box; padding: 0px 5px;">33</li><li style="box-sizing: border-box; padding: 0px 5px;">34</li><li style="box-sizing: border-box; padding: 0px 5px;">35</li><li style="box-sizing: border-box; padding: 0px 5px;">36</li><li style="box-sizing: border-box; padding: 0px 5px;">37</li><li style="box-sizing: border-box; padding: 0px 5px;">38</li><li style="box-sizing: border-box; padding: 0px 5px;">39</li><li style="box-sizing: border-box; padding: 0px 5px;">40</li></ul>

Loss(损失)

<code class="hljs dos has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;">I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728893</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] loss needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728909</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] ip2 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728924</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] relu1 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728938</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] ip1 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728953</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] pool2 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728970</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] conv2 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728984</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] pool1 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">728998</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">170</span>] conv1 needs backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729014</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">172</span>] mnist does <span class="hljs-flow" style="box-sizing: border-box;">not</span> need backward computation.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729027</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">208</span>] This network produces output lossI0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729053</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">467</span>] Collecting Learning Rate and Weight Decay.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729071</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">219</span>] Network initialization done.I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729085</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span>.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">220</span>] Memory required <span class="hljs-flow" style="box-sizing: border-box;">for</span> data: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5169924</span>I0902 <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">13</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">35</span>:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">56</span>.<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">729277</span> <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">16020</span> solver.cpp:<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">156</span>] Creating test <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span> (#<span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">0</span>) specified by <span class="hljs-winutils" style="color: rgb(0, 0, 136); box-sizing: border-box;">net</span> file: examples/mnist/lenet_train_test.prototxt</code><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li></ul><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li><li style="box-sizing: border-box; padding: 0px 5px;">3</li><li style="box-sizing: border-box; padding: 0px 5px;">4</li><li style="box-sizing: border-box; padding: 0px 5px;">5</li><li style="box-sizing: border-box; padding: 0px 5px;">6</li><li style="box-sizing: border-box; padding: 0px 5px;">7</li><li style="box-sizing: border-box; padding: 0px 5px;">8</li><li style="box-sizing: border-box; padding: 0px 5px;">9</li><li style="box-sizing: border-box; padding: 0px 5px;">10</li><li style="box-sizing: border-box; padding: 0px 5px;">11</li><li style="box-sizing: border-box; padding: 0px 5px;">12</li><li style="box-sizing: border-box; padding: 0px 5px;">13</li><li style="box-sizing: border-box; padding: 0px 5px;">14</li></ul>

Completion(完成)

<code class="hljs css has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;"><span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">I0902</span> 13<span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:35</span><span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:56</span><span class="hljs-class" style="box-sizing: border-box; color: rgb(155, 112, 63);">.806970</span> 16020 <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">solver</span><span class="hljs-class" style="box-sizing: border-box; color: rgb(155, 112, 63);">.cpp</span><span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:46</span>] <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">Solver</span> <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">scaffolding</span> <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">done</span>.<span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">I0902</span> 13<span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:35</span><span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:56</span><span class="hljs-class" style="box-sizing: border-box; color: rgb(155, 112, 63);">.806984</span> 16020 <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">solver</span><span class="hljs-class" style="box-sizing: border-box; color: rgb(155, 112, 63);">.cpp</span><span class="hljs-pseudo" style="color: rgb(0, 0, 0); box-sizing: border-box;">:165</span>] <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">Solving</span> <span class="hljs-tag" style="color: rgb(0, 0, 0); box-sizing: border-box;">LeNet</span></code><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li></ul><ul class="pre-numbering" style="box-sizing: border-box; position: absolute; width: 50px; top: 0px; left: 0px; margin: 0px; padding: 6px 0px 40px; border-right-width: 1px; border-right-style: solid; border-right-color: rgb(221, 221, 221); list-style: none; text-align: right; background-color: rgb(238, 238, 238);"><li style="box-sizing: border-box; padding: 0px 5px;">1</li><li style="box-sizing: border-box; padding: 0px 5px;">2</li></ul>

Updating Parameters

权重更新是在Solver::ComputeUpdateValue()中完成的。ComputeUpdateValue方法利用权重衰减项(weight decay)计算出权重梯度(weight gradients)(现阶段还只有误差梯度),进而计算出每个权重最后的梯度。然后经过与学习速率α相乘后,该值被存储于每一个参数(Blob的diff域)中。最后,调用每个参数的Blob::Update()方法进行最后的参数更新(Blob的data值减去diff值)。


Snapshotting and Resuming

Solver使用Solver::Snapshot()方法保存权重的值,使用Solver::SnapshotSolverState()保存状态。保存的权重表示当前学习到的模型,并且允许从该状态继续进行学习。训练可以通过Solver::Restore()方法和Solver::RestoreSolverState()方法继续进行。

权重保存文件没有拓展名,状态保存文件的拓展名是.solverstate extension,每种文件都有一个_iter_N后缀,表明保存时的迭代次数。

在Solver定义文件(*_solver.prototxt)中,保存方法是如下定义:

<code class="hljs vala has-numbering" style="display: block; padding: 0px; color: inherit; box-sizing: border-box; font-family: 'Source Code Pro', monospace;font-size:undefined; white-space: pre; border-radius: 0px; word-wrap: normal; background: transparent;"><span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 保存间隔(单位:迭代次数)</span>snapshot: <span class="hljs-number" style="color: rgb(0, 102, 102); box-sizing: border-box;">5000</span><span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 保存文件的前缀,注意:是相对于caffe根目录,而不是具体Solver定义文件目录</span>snapshot_prefix: <span class="hljs-string" style="color: rgb(0, 136, 0); box-sizing: border-box;">"/path/to/model"</span><span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 是否保存Blob的diff域(存储梯度值),有助于debug,但是耗费更多存储空间</span>snapshot_diff: <span class="hljs-literal" style="color: rgb(0, 102, 102); box-sizing: border-box;">false</span><span class="hljs-preprocessor" style="color: rgb(68, 68, 68); box-sizing: border-box;"># 训练完成后是否保存最后的状态,默认是True</span>snapshot_after_train: <span class="hljs-literal" style="color: rgb(0, 102, 102); box-sizing: border-box;">true</span></code>
0 0
原创粉丝点击