Modified: incubator/singa/site/trunk/en/docs/metric.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/metric.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/metric.html (original) +++ incubator/singa/site/trunk/en/docs/metric.html Wed May 9 15:25:26 2018 @@ -188,8 +188,187 @@ <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article";> <div itemprop="articleBody"> - <div class="section" id="metric"> -<h1>Metric<a class="headerlink" href="#metric" title="Permalink to this headline">¶</a></h1> + <div class="section" id="module-singa.metric"> +<span id="metric"></span><h1>Metric<a class="headerlink" href="#module-singa.metric" title="Permalink to this headline">¶</a></h1> +<p>This module includes a set of metric classes for evaluating the modelâs +performance. The specific metric classes could be converted from C++ +implmentation or implemented directly using Python.</p> +<p>Example usage:</p> +<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">tensor</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">metric</span> + +<span class="n">x</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">Tensor</span><span class="p">((</span><span class="mi">3</span><span class="p">,</span> <span class="mi">5</span><span class="p">))</span> +<span class="n">x</span><span class="o">.</span><span class="n">uniform</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="c1"># randomly genearte the prediction activation</span> +<span class="n">x</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">SoftMax</span><span class="p">(</span><span class="n">x</span><span class="p">)</span> <span class="c1"># normalize the prediction into probabilities</span> +<span class="n">y</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">from_numpy</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">array</span><span class="p">([</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">],</span> <span class="n">dtype</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">int</span><span class="p">))</span> <span class="c1"># set the truth</span> + +<span class="n">f</span> <span class="o">=</span> <span class="n">metric</span><span class="o">.</span><span class="n">Accuracy</span><span class="p">()</span> +<span class="n">acc</span> <span class="o">=</span> <span class="n">f</span><span class="o">.</span><span class="n">evaluate</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">)</span> <span class="c1"># averaged accuracy over all 3 samples in x</span> +</pre></div> +</div> +<dl class="class"> +<dt id="singa.metric.Metric"> +<em class="property">class </em><code class="descclassname">singa.metric.</code><code class="descname">Metric</code><a class="headerlink" href="#singa.metric.Metric" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p> +<p>Base metric class.</p> +<p>Subclasses that wrap the C++ loss classes can use the inherited foward, +and evaluate functions of this base class. Other subclasses need +to override these functions. Users need to feed in the <strong>predictions</strong> and +ground truth to get the metric values.</p> +<dl class="method"> +<dt id="singa.metric.Metric.forward"> +<code class="descname">forward</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Metric.forward" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the metric for each sample.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth values, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tensor of floats, one per sample</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.metric.Metric.evaluate"> +<code class="descname">evaluate</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Metric.evaluate" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the averaged metric over all samples.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth values, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a float value for the averaged metric</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.metric.Accuracy"> +<em class="property">class </em><code class="descclassname">singa.metric.</code><code class="descname">Accuracy</code><a class="headerlink" href="#singa.metric.Accuracy" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.metric.Metric" title="singa.metric.Metric"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.metric.Metric</span></code></a></p> +<p>Compute the top one accuracy for single label prediction tasks.</p> +<p>It calls the C++ functions to do the calculation.</p> +</dd></dl> + +<dl class="class"> +<dt id="singa.metric.Precision"> +<em class="property">class </em><code class="descclassname">singa.metric.</code><code class="descname">Precision</code><span class="sig-paren">(</span><em>top_k</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Precision" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.metric.Metric" title="singa.metric.Metric"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.metric.Metric</span></code></a></p> +<p>Make the top-k labels of max probability as the prediction</p> +<p>Compute the precision against the groundtruth labels</p> +<dl class="method"> +<dt id="singa.metric.Precision.forward"> +<code class="descname">forward</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Precision.forward" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the precision for each sample.</p> +<p>Convert tensor to numpy for computation</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth labels, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tensor of floats, one per sample</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.metric.Precision.evaluate"> +<code class="descname">evaluate</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Precision.evaluate" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the averaged precision over all samples.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth values, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a float value for the averaged metric</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.metric.Recall"> +<em class="property">class </em><code class="descclassname">singa.metric.</code><code class="descname">Recall</code><span class="sig-paren">(</span><em>top_k</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Recall" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.metric.Metric" title="singa.metric.Metric"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.metric.Metric</span></code></a></p> +<p>Make the top-k labels of max probability as the prediction</p> +<p>Compute the recall against the groundtruth labels</p> +<dl class="method"> +<dt id="singa.metric.Recall.forward"> +<code class="descname">forward</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Recall.forward" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the recall for each sample.</p> +<p>Convert tensor to numpy for computation</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth labels, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tensor of floats, one per sample</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.metric.Recall.evaluate"> +<code class="descname">evaluate</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.metric.Recall.evaluate" title="Permalink to this definition">¶</a></dt> +<dd><p>Compute the averaged precision over all samples.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â predictions, one row per sample</li> +<li><strong>y</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â ground truth values, one row per sample</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a float value for the averaged metric</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + </div> @@ -255,9 +434,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script>
Modified: incubator/singa/site/trunk/en/docs/model_zoo/caffe/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/caffe/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/caffe/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/caffe/README.html Wed May 9 15:25:26 2018 @@ -259,9 +259,7 @@ You can start the prediction program by <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/char-rnn/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/char-rnn/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/char-rnn/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/char-rnn/README.html Wed May 9 15:25:26 2018 @@ -299,9 +299,7 @@ Other plain text files can also be used. <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/cifar10/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/cifar10/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/cifar10/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/cifar10/README.html Wed May 9 15:25:26 2018 @@ -344,9 +344,7 @@ The âmodel.binâ file generated <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/imagenet/alexnet/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/imagenet/alexnet/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/imagenet/alexnet/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/imagenet/alexnet/README.html Wed May 9 15:25:26 2018 @@ -339,9 +339,7 @@ folder in data preprocessing step;</li> <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/imagenet/googlenet/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/imagenet/googlenet/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/imagenet/googlenet/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/imagenet/googlenet/README.html Wed May 9 15:25:26 2018 @@ -334,9 +334,7 @@ Refer to <a class="reference external" h <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/imagenet/inception/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/imagenet/inception/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/imagenet/inception/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/imagenet/inception/README.html Wed May 9 15:25:26 2018 @@ -270,9 +270,7 @@ After downloading and decompressing the <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/imagenet/resnet/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/imagenet/resnet/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/imagenet/resnet/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/imagenet/resnet/README.html Wed May 9 15:25:26 2018 @@ -281,9 +281,7 @@ the convert.py program.</p> <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/imagenet/vgg/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/imagenet/vgg/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/imagenet/vgg/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/imagenet/vgg/README.html Wed May 9 15:25:26 2018 @@ -278,9 +278,7 @@ to SINGA for image classification.</p> <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/index.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/index.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/index.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/index.html Wed May 9 15:25:26 2018 @@ -294,9 +294,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/model_zoo/mnist/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/model_zoo/mnist/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/model_zoo/mnist/README.html (original) +++ incubator/singa/site/trunk/en/docs/model_zoo/mnist/README.html Wed May 9 15:25:26 2018 @@ -244,9 +244,7 @@ the program with an additional argument< <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/net.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/net.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/net.html (original) +++ incubator/singa/site/trunk/en/docs/net.html Wed May 9 15:25:26 2018 @@ -188,8 +188,298 @@ <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article";> <div itemprop="articleBody"> - <div class="section" id="feedforward-net"> -<h1>FeedForward Net<a class="headerlink" href="#feedforward-net" title="Permalink to this headline">¶</a></h1> + <div class="section" id="module-singa.net"> +<span id="feedforward-net"></span><h1>FeedForward Net<a class="headerlink" href="#module-singa.net" title="Permalink to this headline">¶</a></h1> +<p>Nerual net class for constructing the nets using layers and providing access +functions for net info, e.g., parameters.</p> +<p>Example usages:</p> +<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">net</span> <span class="k">as</span> <span class="n">ffnet</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">metric</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">loss</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">layer</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">device</span> + +<span class="c1"># create net and add layers</span> +<span class="n">net</span> <span class="o">=</span> <span class="n">ffnet</span><span class="o">.</span><span class="n">FeedForwardNet</span><span class="p">(</span><span class="n">loss</span><span class="o">.</span><span class="n">SoftmaxCrossEntropy</span><span class="p">(),</span> <span class="n">metric</span><span class="o">.</span><span class="n">Accuracy</span><span class="p">())</span> +<span class="n">net</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">layer</span><span class="o">.</span><span class="n">Conv2D</span><span class="p">(</span><span class="s1">'conv1'</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">5</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">input_sample_shape</span><span class="o">=</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span><span class="mi">32</span><span class="p">,</span><span class="mi">32</span><span class="p">,)))</span> +<span class="n">net</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">layer</span><span class="o">.</span><span class="n">Activation</span><span class="p">(</span><span class="s1">'relu1'</span><span class="p">))</span> +<span class="n">net</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">layer</span><span class="o">.</span><span class="n">MaxPooling2D</span><span class="p">(</span><span class="s1">'pool1'</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span> +<span class="n">net</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">layer</span><span class="o">.</span><span class="n">Flatten</span><span class="p">(</span><span class="s1">'flat'</span><span class="p">))</span> +<span class="n">net</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">layer</span><span class="o">.</span><span class="n">Dense</span><span class="p">(</span><span class="s1">'dense'</span><span class="p">,</span> <span class="mi">10</span><span class="p">))</span> + +<span class="c1"># init parameters</span> +<span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">net</span><span class="o">.</span><span class="n">param_values</span><span class="p">():</span> + <span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">shape</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span> + <span class="n">p</span><span class="o">.</span><span class="n">set_value</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> + <span class="k">else</span><span class="p">:</span> + <span class="n">p</span><span class="o">.</span><span class="n">gaussian</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.01</span><span class="p">)</span> + +<span class="c1"># move net onto gpu</span> +<span class="n">dev</span> <span class="o">=</span> <span class="n">device</span><span class="o">.</span><span class="n">create_cuda_gpu</span><span class="p">()</span> +<span class="n">net</span><span class="o">.</span><span class="n">to_device</span><span class="p">(</span><span class="n">dev</span><span class="p">)</span> + +<span class="c1"># training (skipped)</span> + +<span class="c1"># do prediction after training</span> +<span class="n">x</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">Tensor</span><span class="p">((</span><span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">32</span><span class="p">),</span> <span class="n">dev</span><span class="p">)</span> +<span class="n">x</span><span class="o">.</span><span class="n">uniform</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> +<span class="n">y</span> <span class="o">=</span> <span class="n">net</span><span class="o">.</span><span class="n">predict</span><span class="p">(</span><span class="n">x</span><span class="p">)</span> +<span class="nb">print</span> <span class="n">tensor</span><span class="o">.</span><span class="n">to_numpy</span><span class="p">(</span><span class="n">y</span><span class="p">)</span> +</pre></div> +</div> +<dl class="class"> +<dt id="singa.net.FeedForwardNet"> +<em class="property">class </em><code class="descclassname">singa.net.</code><code class="descname">FeedForwardNet</code><span class="sig-paren">(</span><em>loss=None</em>, <em>metric=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p> +<dl class="method"> +<dt id="singa.net.FeedForwardNet.to_device"> +<code class="descname">to_device</code><span class="sig-paren">(</span><em>dev</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.to_device" title="Permalink to this definition">¶</a></dt> +<dd><p>Move the net onto the given device, including +all parameters and intermediate data.</p> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.add"> +<code class="descname">add</code><span class="sig-paren">(</span><em>lyr</em>, <em>src=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.add" title="Permalink to this definition">¶</a></dt> +<dd><p>Append a layer into the layer list.</p> +<p>This function will get the sample shape from the src layers to setup the +newly added layer. For the first layer, it is setup outside. The calling +function should ensure the correctness of the layer order. If src is +None, the last layer is the src layer. If there are multiple src layers, +the src is a list of the src layers.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>lyr</strong> (<a class="reference internal" href="layer.html#singa.layer.Layer" title="singa.layer.Layer"><em>Layer</em></a>) â the layer to be added</li> +<li><strong>src</strong> (<a class="reference internal" href="layer.html#singa.layer.Layer" title="singa.layer.Layer"><em>Layer</em></a>) â the source layer of lyr</li> +</ul> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.param_values"> +<code class="descname">param_values</code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.param_values" title="Permalink to this definition">¶</a></dt> +<dd><p>Return a list of tensors for all parameters</p> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.param_specs"> +<code class="descname">param_specs</code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.param_specs" title="Permalink to this definition">¶</a></dt> +<dd><p>Return a list of ParamSpec for all parameters</p> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.param_names"> +<code class="descname">param_names</code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.param_names" title="Permalink to this definition">¶</a></dt> +<dd><p>Return a list for the names of all params</p> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.train"> +<code class="descname">train</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.train" title="Permalink to this definition">¶</a></dt> +<dd><p>Run BP for one iteration. +This method is deprecated. It is only kept for backward compatibility. +The name of this method is confusing since it does not update parameters. +Please use backprob() instead. +The back progagation algorithm computes gradients but it does not train.</p> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.backprob"> +<code class="descname">backprob</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.backprob" title="Permalink to this definition">¶</a></dt> +<dd><p>Run BP for one iteration.</p> +<p>Currently only support nets with a single output layer, and a single +loss objective and metric. +For multiple outputs (with multiple loss/metric), please manually +call forward, compute loss/metric and call backward. backward() is also +more memory efficient than this function.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>x</strong> â input data, a single input Tensor or a dict: layer name -> Tensor</li> +<li><strong>y</strong> â label data, a single input Tensor.</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">gradients of parameters and the loss and metric values.</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.evaluate"> +<code class="descname">evaluate</code><span class="sig-paren">(</span><em>x</em>, <em>y</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.evaluate" title="Permalink to this definition">¶</a></dt> +<dd><p>Evaluate the loss and metric of the given data.</p> +<p>Currently only support nets with a single output layer, and a single +loss objective and metric. +TODO(wangwei) consider multiple loss objectives and metrics.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>x</strong> â input data, a single input Tensor or a dict: layer name -> Tensor</li> +<li><strong>y</strong> â label data, a single input Tensor.</li> +</ul> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.predict"> +<code class="descname">predict</code><span class="sig-paren">(</span><em>x</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.predict" title="Permalink to this definition">¶</a></dt> +<dd><p>Forward the input data through each layer to get the values of the +output layers.</p> +<p>Currently only support nets with a single output layer</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>x</strong> â input data, a single input Tensor or a dict: layer name -> Tensor</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body">a single output tensor as the prediction result.</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.topo_sort"> +<code class="descname">topo_sort</code><span class="sig-paren">(</span><em>layers</em>, <em>src_of_layer</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.topo_sort" title="Permalink to this definition">¶</a></dt> +<dd><p>Topology sort of layers.</p> +<p>It would try to preserve the orders of the input layers.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>layers</strong> â a list of layers; the layers from the output of the same +layer (e.g., slice layer) should be added by users in correct +order; This function would not change their order.</li> +<li><strong>src_of_layer</strong> â a dictionary: src layer name -> a list of src layers</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">A list of ordered layer</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.forward"> +<code class="descname">forward</code><span class="sig-paren">(</span><em>flag</em>, <em>x</em>, <em>output=[]</em>, <em>freeze=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.forward" title="Permalink to this definition">¶</a></dt> +<dd><p>Forward the input(s) through every layer.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>flag</strong> â True for training; False for evaluation; could also be +model_pb2.kTrain or model_pb2.kEval, or other values for future +use.</li> +<li><strong>x</strong> â a single SINGA tensor if there is a single input; otherwise, a +dictionary: layer name-> singa tensor, for each layer accepting +input data. Do not associate a layer with input tensor if it is +connected from another layer. For such case, use a Dummy() layer +to accept the input data and connect the dummy layer to this +layer.</li> +<li><strong>output</strong> (<em>list</em>) â a list of layer names whose output would be returned +in addition to the default output.</li> +<li><strong>freeze</strong> (<em>str</em>) â layer name, freeze all layers before this layer; flag +is set to false for these layers.</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last"><dl class="docutils"> +<dt>if there is only one output layer and output arg is empty, return</dt> +<dd><p class="first last">the result from the single output layer; otherwise, return a +dictionary: layer name -> output tensor(s)</p> +</dd> +</dl> +</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.backward"> +<code class="descname">backward</code><span class="sig-paren">(</span><em>dy</em>, <em>output=[]</em>, <em>freeze=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.backward" title="Permalink to this definition">¶</a></dt> +<dd><p>Run back-propagation after forward-propagation.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>dy</strong> â a single tensor if there is a single loss function; otherwise, +a dictionary maps the name of the layer connecting to the loss +function -> gradient from the loss function. Do not associate a +layer with gradient tensor if it is connecting to another layer. +For such case, connect this layer to a Dummy() layer and use the +dummy layer to accept the gradient.</li> +<li><strong>output</strong> (<em>list</em>) â a list of layer names whose output gradient would be +returned in addition to the param gradient</li> +<li><strong>freeze</strong> (<em>str</em>) â layer name, stop backward after this layer.</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a geneartor iterator that generates +(param_names, param_values, param_grads, layer_grads) after +processing each layer h, where the first three lists are for h +and the last item is a dictionary which maps +layer name -> its output gradient tensor(s). At the end of this +function, the key set includes all layers in the output arg.</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.save"> +<code class="descname">save</code><span class="sig-paren">(</span><em>f</em>, <em>buffer_size=10</em>, <em>use_pickle=False</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.save" title="Permalink to this definition">¶</a></dt> +<dd><p>Save model parameters using io/snapshot.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>f</strong> â file name</li> +<li><strong>buffer_size</strong> â size (MB) of the IO, default setting is 10MB; Please +make sure it is larger than any single parameter object.</li> +<li><strong>use_pickle</strong> (<em>Boolean</em>) â if true, it would use pickle for dumping; +otherwise, it would use protobuf for serialization, which uses +less space.</li> +</ul> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.net.FeedForwardNet.load"> +<code class="descname">load</code><span class="sig-paren">(</span><em>f</em>, <em>buffer_size=10</em>, <em>use_pickle=False</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.net.FeedForwardNet.load" title="Permalink to this definition">¶</a></dt> +<dd><p>Load model parameters using io/snapshot.</p> +<p>Please refer to the argument description in save().</p> +</dd></dl> + +</dd></dl> + </div> @@ -255,9 +545,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/neural-net.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/neural-net.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/neural-net.html (original) +++ incubator/singa/site/trunk/en/docs/neural-net.html Wed May 9 15:25:26 2018 @@ -544,9 +544,7 @@ lower layers and <code class="docutils l <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/notebook/README.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/notebook/README.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/notebook/README.html (original) +++ incubator/singa/site/trunk/en/docs/notebook/README.html Wed May 9 15:25:26 2018 @@ -223,9 +223,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/optimizer.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/optimizer.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/optimizer.html (original) +++ incubator/singa/site/trunk/en/docs/optimizer.html Wed May 9 15:25:26 2018 @@ -188,8 +188,462 @@ <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article";> <div itemprop="articleBody"> - <div class="section" id="optimizer"> -<h1>Optimizer<a class="headerlink" href="#optimizer" title="Permalink to this headline">¶</a></h1> + <div class="section" id="module-singa.optimizer"> +<span id="optimizer"></span><h1>Optimizer<a class="headerlink" href="#module-singa.optimizer" title="Permalink to this headline">¶</a></h1> +<p>This module includes a set of optimizers for updating model parameters.</p> +<p>Example usage:</p> +<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">optimizer</span> +<span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">tensor</span> + +<span class="n">sgd</span> <span class="o">=</span> <span class="n">optimizer</span><span class="o">.</span><span class="n">SGD</span><span class="p">(</span><span class="n">lr</span><span class="o">=</span><span class="mf">0.01</span><span class="p">,</span> <span class="n">momentum</span><span class="o">=</span><span class="mf">0.9</span><span class="p">,</span> <span class="n">weight_decay</span><span class="o">=</span><span class="mf">1e-4</span><span class="p">)</span> +<span class="n">p</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">Tensor</span><span class="p">((</span><span class="mi">3</span><span class="p">,</span><span class="mi">5</span><span class="p">))</span> +<span class="n">p</span><span class="o">.</span><span class="n">uniform</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> +<span class="n">g</span> <span class="o">=</span> <span class="n">tensor</span><span class="o">.</span><span class="n">Tensor</span><span class="p">((</span><span class="mi">3</span><span class="p">,</span><span class="mi">5</span><span class="p">))</span> +<span class="n">g</span><span class="o">.</span><span class="n">gaussian</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mf">0.01</span><span class="p">)</span> + +<span class="n">sgd</span><span class="o">.</span><span class="n">apply</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">g</span><span class="p">,</span> <span class="n">p</span><span class="p">,</span> <span class="s1">'param'</span><span class="p">)</span> <span class="c1"># use the global lr=0.1 for epoch 1</span> +<span class="n">sgd</span><span class="o">.</span><span class="n">apply_with_lr</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mf">0.03</span><span class="p">,</span> <span class="n">g</span><span class="p">,</span> <span class="n">p</span><span class="p">,</span> <span class="s1">'param'</span><span class="p">)</span> <span class="c1"># use lr=0.03 for epoch 2</span> +</pre></div> +</div> +<dl class="class"> +<dt id="singa.optimizer.Optimizer"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">Optimizer</code><span class="sig-paren">(</span><em>lr=None</em>, <em>momentum=None</em>, <em>weight_decay=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Optimizer" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p> +<p>The base python optimizer class.</p> +<p>Typically, an optimizer is used as follows:</p> +<ol class="arabic simple"> +<li>construct the optimizer</li> +<li>(optional) register each parameter with its specs.</li> +<li>use the optimizer to update parameter values given parameter gradients +and other optional info</li> +</ol> +<p>The subclasses should override the apply_with_lr function to do the real +parameter udpate.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>lr</strong> (<em>float</em>) â a constant value for the learning rate</li> +<li><strong>momentum</strong> (<em>float</em>) â a constant value for the momentum value</li> +<li><strong>weight_decay</strong> (<em>float</em>) â the coefficent for L2 regularizer, which is +mutually exclusive with âregularizerâ.</li> +<li><strong>regularizer</strong> â an instance of Regularizer or RegularizerConf; If set, +regularization would be applied in apply_with_lr(). +Users can also do regularization outside.</li> +<li><strong>constraint</strong> â an instance of Constraint or ConstraintConf; If set, +constraint would be applied inside apply_with_lr(). Users can +also apply constraint outside.</li> +</ul> +</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.Optimizer.register"> +<code class="descname">register</code><span class="sig-paren">(</span><em>name</em>, <em>specs</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Optimizer.register" title="Permalink to this definition">¶</a></dt> +<dd><p>Register the param specs, including creating regularizer and +constraint per param object. Param specific regularizer and constraint +have higher priority than the global ones. If all parameters share the +same setting for learning rate, regularizer and constraint, then there +is no need to call this function.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>name</strong> (<em>str</em>) â parameter name</li> +<li><strong>specs</strong> (<em>ParamSpec</em>) â protobuf obj, including regularizer and +constraint, multipliers for learning rate and weight decay.</li> +</ul> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.optimizer.Optimizer.apply_regularizer_constraint"> +<code class="descname">apply_regularizer_constraint</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>name=None</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Optimizer.apply_regularizer_constraint" title="Permalink to this definition">¶</a></dt> +<dd><p>Apply regularization and constraint if available.</p> +<p>If there are both global regularizer (constraint) and param specific +regularizer (constraint), it would use the param specific one.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>value</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter value Tensor</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient Tensor</li> +<li><strong>name</strong> (<em>string</em>) â to get parameter specific regularizer or constraint</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">the updated gradient Tensor</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.optimizer.Optimizer.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name=None</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Optimizer.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update of parameters with given learning rate if the grad is not +empty.</p> +<p>The subclass optimizer must override this function. +This function do nothing if the grad is empty.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>lr</strong> (<em>float</em>) â learning rate</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to index parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.optimizer.Optimizer.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>grad</em>, <em>value</em>, <em>name=None</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Optimizer.apply" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update assuming the learning rate generator is set.</p> +<p>The subclass optimizer does not need to override this function.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to retrieval parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â training iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.SGD"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">SGD</code><span class="sig-paren">(</span><em>lr=None</em>, <em>momentum=None</em>, <em>weight_decay=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.SGD" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Optimizer" title="singa.optimizer.Optimizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Optimizer</span></code></a></p> +<p>The vallina Stochasitc Gradient Descent algorithm with momentum.</p> +<p>See the base Optimizer for all arguments.</p> +<dl class="method"> +<dt id="singa.optimizer.SGD.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.SGD.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update of parameters with given learning rate if the grad is not +empty.</p> +<p>The subclass optimizer must override this function. +This function do nothing if the grad is empty.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>lr</strong> (<em>float</em>) â learning rate</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to index parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.Nesterov"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">Nesterov</code><span class="sig-paren">(</span><em>lr=None</em>, <em>momentum=0.9</em>, <em>weight_decay=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Nesterov" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Optimizer" title="singa.optimizer.Optimizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Optimizer</span></code></a></p> +<p>The SGD with Nesterov momentum.</p> +<p>See the base Optimizer for all arguments.</p> +<dl class="method"> +<dt id="singa.optimizer.Nesterov.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Nesterov.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update of parameters with given learning rate if the grad is not +empty.</p> +<p>The subclass optimizer must override this function. +This function do nothing if the grad is empty.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>lr</strong> (<em>float</em>) â learning rate</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to index parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.RMSProp"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">RMSProp</code><span class="sig-paren">(</span><em>rho=0.9</em>, <em>epsilon=1e-08</em>, <em>lr=None</em>, <em>weight_decay=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.RMSProp" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Optimizer" title="singa.optimizer.Optimizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Optimizer</span></code></a></p> +<p>RMSProp optimizer.</p> +<p>See the base Optimizer for all constructor args.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>rho</strong> (<em>float</em>) â float within [0, 1]</li> +<li><strong>epsilon</strong> (<em>float</em>) â small value for preventing numeric error</li> +</ul> +</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.RMSProp.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.RMSProp.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update of parameters with given learning rate if the grad is not +empty.</p> +<p>The subclass optimizer must override this function. +This function do nothing if the grad is empty.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>lr</strong> (<em>float</em>) â learning rate</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to index parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.AdaGrad"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">AdaGrad</code><span class="sig-paren">(</span><em>epsilon=1e-08</em>, <em>lr=None</em>, <em>weight_decay=None</em>, <em>lr_gen=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.AdaGrad" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Optimizer" title="singa.optimizer.Optimizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Optimizer</span></code></a></p> +<p>AdaGrad optimizer.</p> +<p>See the base Optimizer for all constructor args.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>epsilon</strong> (<em>float</em>) â small number for preventing numeric error.</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.AdaGrad.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.AdaGrad.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Do update of parameters with given learning rate if the grad is not +empty.</p> +<p>The subclass optimizer must override this function. +This function do nothing if the grad is empty.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple"> +<li><strong>epoch</strong> (<em>int</em>) â training epoch ID</li> +<li><strong>lr</strong> (<em>float</em>) â learning rate</li> +<li><strong>grad</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â parameter gradient</li> +<li><strong>value</strong> (<em>Tesnor</em>) â parameter value</li> +<li><strong>name</strong> (<em>string</em>) â paramter name to index parameter specific +updating rules (including regularizer and constraint)</li> +<li><strong>step</strong> (<em>int</em>) â iteration ID within one epoch</li> +</ul> +</td> +</tr> +<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">updated parameter value</p> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.Adam"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">Adam</code><span class="sig-paren">(</span><em>beta_1=0.9</em>, <em>beta_2=0.999</em>, <em>epsilon=1e-08</em>, <em>lr=None</em>, <em>weight_decay=None</em>, <em>regularizer=None</em>, <em>constraint=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Adam" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Optimizer" title="singa.optimizer.Optimizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Optimizer</span></code></a></p> +<p>Adam optimizer.</p> +<p>See the base Optimizer for all constructor args.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>beta_1</strong> (<em>float</em>) â coefficient of momentum</li> +<li><strong>beta_2</strong> (<em>float</em>) â coefficient of aggregated squared gradient</li> +<li><strong>epsilon</strong> (<em>float</em>) â small value for preventing numeric error</li> +</ul> +</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.Adam.apply_with_lr"> +<code class="descname">apply_with_lr</code><span class="sig-paren">(</span><em>epoch</em>, <em>lr</em>, <em>grad</em>, <em>value</em>, <em>name</em>, <em>step</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Adam.apply_with_lr" title="Permalink to this definition">¶</a></dt> +<dd><p>Update one parameter object.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>step</strong> (<em>int</em>) â the accumulated training iterations, not the iteration ID</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.Regularizer"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">Regularizer</code><a class="headerlink" href="#singa.optimizer.Regularizer" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p> +<p>Base Python regularizer for parameter gradients.</p> +<dl class="method"> +<dt id="singa.optimizer.Regularizer.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Regularizer.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.CppRegularizer"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">CppRegularizer</code><span class="sig-paren">(</span><em>conf</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.CppRegularizer" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Regularizer" title="singa.optimizer.Regularizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Regularizer</span></code></a></p> +<p>Wrapper for regularizer implemented using C++.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>conf</strong> (<em>RegularizerConf</em>) â protobuf message for the configuration.</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.CppRegularizer.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.CppRegularizer.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.L2Regularizer"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">L2Regularizer</code><span class="sig-paren">(</span><em>coefficient</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.L2Regularizer" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Regularizer" title="singa.optimizer.Regularizer"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Regularizer</span></code></a></p> +<p>L2 regularization</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>coefficient</strong> (<em>float</em>) â regularization coefficient.</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.L2Regularizer.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.L2Regularizer.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.Constraint"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">Constraint</code><a class="headerlink" href="#singa.optimizer.Constraint" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p> +<p>Base Python constraint class for paramter gradients</p> +<dl class="method"> +<dt id="singa.optimizer.Constraint.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.Constraint.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.CppConstraint"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">CppConstraint</code><span class="sig-paren">(</span><em>conf</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.CppConstraint" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Constraint" title="singa.optimizer.Constraint"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Constraint</span></code></a></p> +<p>Wrapper for constraints implemented using C++.</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><strong>conf</strong> (<em>ConstraintConf</em>) â protobuf message for the configuration.</td> +</tr> +</tbody> +</table> +<dl class="method"> +<dt id="singa.optimizer.CppConstraint.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.CppConstraint.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + +<dl class="class"> +<dt id="singa.optimizer.L2Constraint"> +<em class="property">class </em><code class="descclassname">singa.optimizer.</code><code class="descname">L2Constraint</code><span class="sig-paren">(</span><em>threshold=None</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.L2Constraint" title="Permalink to this definition">¶</a></dt> +<dd><p>Bases: <a class="reference internal" href="#singa.optimizer.Constraint" title="singa.optimizer.Constraint"><code class="xref py py-class docutils literal notranslate"><span class="pre">singa.optimizer.Constraint</span></code></a></p> +<p>Rescale the gradient to make the L2 norm <= a given threshold</p> +<dl class="method"> +<dt id="singa.optimizer.L2Constraint.apply"> +<code class="descname">apply</code><span class="sig-paren">(</span><em>epoch</em>, <em>value</em>, <em>grad</em>, <em>step=-1</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.optimizer.L2Constraint.apply" title="Permalink to this definition">¶</a></dt> +<dd></dd></dl> + +</dd></dl> + </div> @@ -255,9 +709,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/snapshot.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/snapshot.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/snapshot.html (original) +++ incubator/singa/site/trunk/en/docs/snapshot.html Wed May 9 15:25:26 2018 @@ -188,8 +188,58 @@ <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article";> <div itemprop="articleBody"> - <div class="section" id="snapshot"> -<h1>Snapshot<a class="headerlink" href="#snapshot" title="Permalink to this headline">¶</a></h1> + <div class="section" id="module-singa.snapshot"> +<span id="snapshot"></span><h1>Snapshot<a class="headerlink" href="#module-singa.snapshot" title="Permalink to this headline">¶</a></h1> +<p>This script includes io::snapshot class and its methods.</p> +<p>Example usages:</p> +<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">singa</span> <span class="k">import</span> <span class="n">snapshot</span> + +<span class="n">sn1</span> <span class="o">=</span> <span class="n">snapshot</span><span class="o">.</span><span class="n">Snapshot</span><span class="p">(</span><span class="s1">'param'</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span> +<span class="n">params</span> <span class="o">=</span> <span class="n">sn1</span><span class="o">.</span><span class="n">read</span><span class="p">()</span> <span class="c1"># read all params as a dictionary</span> + +<span class="n">sn2</span> <span class="o">=</span> <span class="n">snapshot</span><span class="o">.</span><span class="n">Snapshot</span><span class="p">(</span><span class="s1">'param_new'</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span> +<span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">params</span><span class="o">.</span><span class="n">iteritems</span><span class="p">():</span> + <span class="n">sn2</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">k</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span> +</pre></div> +</div> +<dl class="class"> +<dt id="singa.snapshot.Snapshot"> +<em class="property">class </em><code class="descclassname">singa.snapshot.</code><code class="descname">Snapshot</code><span class="sig-paren">(</span><em>f</em>, <em>mode</em>, <em>buffer_size=10</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.snapshot.Snapshot" title="Permalink to this definition">¶</a></dt> +<dd><p>Class and member functions for singa::Snapshot.</p> +<dl class="method"> +<dt id="singa.snapshot.Snapshot.read"> +<code class="descname">read</code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#singa.snapshot.Snapshot.read" title="Permalink to this definition">¶</a></dt> +<dd><p>Call read method to load all (param_name, param_val)</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Returns:</th><td class="field-body">a dict of (parameter name, parameter Tensor)</td> +</tr> +</tbody> +</table> +</dd></dl> + +<dl class="method"> +<dt id="singa.snapshot.Snapshot.write"> +<code class="descname">write</code><span class="sig-paren">(</span><em>param_name</em>, <em>param_val</em><span class="sig-paren">)</span><a class="headerlink" href="#singa.snapshot.Snapshot.write" title="Permalink to this definition">¶</a></dt> +<dd><p>Call Write method to write a parameter</p> +<table class="docutils field-list" frame="void" rules="none"> +<col class="field-name" /> +<col class="field-body" /> +<tbody valign="top"> +<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple"> +<li><strong>param_name</strong> (<em>string</em>) â name of the parameter</li> +<li><strong>param_val</strong> (<a class="reference internal" href="tensor.html#singa.tensor.Tensor" title="singa.tensor.Tensor"><em>Tensor</em></a>) â value tensor of the parameter</li> +</ul> +</td> +</tr> +</tbody> +</table> +</dd></dl> + +</dd></dl> + </div> @@ -255,9 +305,7 @@ <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script> Modified: incubator/singa/site/trunk/en/docs/software_stack.html URL: http://svn.apache.org/viewvc/incubator/singa/site/trunk/en/docs/software_stack.html?rev=1831260&r1=1831259&r2=1831260&view=diff ============================================================================== --- incubator/singa/site/trunk/en/docs/software_stack.html (original) +++ incubator/singa/site/trunk/en/docs/software_stack.html Wed May 9 15:25:26 2018 @@ -346,9 +346,7 @@ model parameter values using parameter g <script type="text/javascript"> jQuery(function () { - - SphinxRtdTheme.Navigation.enableSticky(); - + SphinxRtdTheme.Navigation.enable(true); }); </script>
