Deci-AI
/
super-gradients
connected to https://github.com/Deci-AI/super-gradients.git


  
1

	
2

	
3

	
4

	
5

	
6

	
7

	
8

	
9

	
10

	
11

	
12

	
13

	
14

	
15

	
16

	
17

	
18

	
19

	
20

	
21

	
22

	
23

	
24

	
25

	
26

	
27

	
28

	
29

	
30

	
31

	
32

	
33

	
34

	
35

	
36

	
37

	
38

	
39

	
40

	
41

	
42

	
43

	
44

	
45

	
46

	
47

	
48

	
49

	
50

	
51

	
52

	
53

	
54

	
55

	
56

	
57

	
58

	
59

	
60

	
61

	
62

	
63

	
64

	
65

	
66

	
67

	
68

	
69

	
70

	
71

	
72

	
73

	
74

	
75

	
76

	
77

	
78

	
79

	
80

	
81

	
82

	
83

	
84

	
85

	
86

	
87

	
88

	
89

	
90

	
91

	
92

	
93

	
94

	
95

	
96

	
97

	
98

	
99

	
100

	
101

	
102

	
103

	
104

	
105

	
106

	
107

	
108

	
109

	
110

	
111

	
112

	
113

	
114

	
115

	
116

	
117

	
118

	
119

	
120

	
121

	
122

	
123

	
124

	
125

	
126

	
127

	
128

	
129

	
130

	
131

	
132

	
133

	
134

	
135

	
136

	
137

	
138

	
139

	
140

	
141

	
142

	
143

	
144

	
145

	
146

	
147

	
148

	
149

	
150

	
151

	
152

	
153

	
154

	
155

	
156

	
157

	
158

	
159

	
160

	
161

	
162

	
163

	
164

	
165

	
166

	
167

	
168

	
169

	
170

	
171

	
172

	
173

	
174

	
175

	
176

	
177

	
178

	
179

	
180

	
181

	
182

	
183

	
184

	
185

	
186

	
187

	
188

	
189

	
190

	
191

	
192

	
193

	
194

	
195

	
196

	
197

	
198

	
199

	
200

	
201

	
202

	
203

	
204

	
205

	
206

	
207

	
208

	
209

	
210

	
211

	
212

	
213

	
214

	
215

	
216

	
217

	
218

	
219

	
220

	
221

	
222

	
223

	
224

	
225

	
226

	
227

	
228

	
229

	
230

	
231

	
232

	
233

	
234

	
235

	
236

	
237

	
238

	
239

	
240

	
241

	
242

	
243

	
244

	
245

	
246

	
247

	
248

	
249

	
250

	
251

	
252

	
253

	
254

	
255

	
256

	
257

	
258

	
259

	
260

	
261

	
262

	
263

	
264

	
265

	
266

	
267

	
268

	
269

	
270

	
271

	
272

	
273

	
274

	
275

	
276

	
277

	
278

	
279

	
280

	
281

	
282

	
283

	
284

	
285

	
286

	
287

	
288

	
289

	
290

	
291

	
292

	
293

	
294

	
295

	
296

	
297

	
298

	
299

	
300

	
301

	
302

	
303

	
304

	
305

	
306

	
307

	
308

	
309

	
310

	
311

	
312

	
313

	
314

	
315

	
316

	
317

	
318

	
319

	
320

	
321

	
322

	
323

	
324

	
325

	
326

	
327

	
328

	
329

	
330

	
331

	
332

	
333

	
334

	
335

	
336

	
337

	
338

	
339

	
340

	
341

	
342

	
343

	
344

	
345

	
346

	
347

	
348

	
349

	
350

	
351

	
352

	
353

	
354

	
355

	
356

	
357

	
358

	
359

	
360

	
361

	
362

	
363

	
364

	
365

	
366

	
367

	
368

	
369

	
370

	
371

	
372

	
373

	
374

	
375

	
376

	
377

	
378

	
379

	
380

	
381

	
382

	
383

	
384

	
385

	
386

	
387

	
388

	
389

	
390

	
391

	
392

	
393

	
            <!DOCTYPE html>
<html class="writer-html5" lang="en" >
<head>
  <meta charset="utf-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
  <title>super_gradients.training.kd_model.kd_model &mdash; SuperGradients 1.0 documentation</title>
      <link rel="stylesheet" href="../../../../_static/pygments.css" type="text/css" />
      <link rel="stylesheet" href="../../../../_static/css/theme.css" type="text/css" />
      <link rel="stylesheet" href="../../../../_static/graphviz.css" type="text/css" />
  <!--[if lt IE 9]>
    <script src="../../../../_static/js/html5shiv.min.js"></script>
  <![endif]-->
  
        <script data-url_root="../../../../" id="documentation_options" src="../../../../_static/documentation_options.js"></script>
        <script src="../../../../_static/jquery.js"></script>
        <script src="../../../../_static/underscore.js"></script>
        <script src="../../../../_static/doctools.js"></script>
    <script src="../../../../_static/js/theme.js"></script>
    <link rel="index" title="Index" href="../../../../genindex.html" />
    <link rel="search" title="Search" href="../../../../search.html" /> 
</head>

<body class="wy-body-for-nav"> 
  <div class="wy-grid-for-nav">
    <nav data-toggle="wy-nav-shift" class="wy-nav-side">
      <div class="wy-side-scroll">
        <div class="wy-side-nav-search" >
            <a href="../../../../index.html" class="icon icon-home"> SuperGradients
          </a>
<div role="search">
  <form id="rtd-search-form" class="wy-form" action="../../../../search.html" method="get">
    <input type="text" name="q" placeholder="Search docs" />
    <input type="hidden" name="check_keywords" value="yes" />
    <input type="hidden" name="area" value="default" />
  </form>
</div>
        </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
              <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html">Fill our 4-question quick survey! We will raffle free SuperGradients swag between those who will participate -&gt; Fill Survey</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html#supergradients">SuperGradients</a></li>
</ul>
<p class="caption"><span class="caption-text">Technical Documentation</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.common.html">Common package</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.training.html">Training package</a></li>
</ul>
<p class="caption"><span class="caption-text">User Guide</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html">What is SuperGradients?</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#introducing-the-supergradients-library">Introducing the SuperGradients library</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#installation">Installation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#integrating-your-training-code-complete-walkthrough">Integrating Your Training Code - Complete Walkthrough</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#training-parameters">Training Parameters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#logs-and-checkpoints">Logs and Checkpoints</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#dataset-parameters">Dataset Parameters</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#network-architectures">Network Architectures</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#pretrained-models">Pretrained Models</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#how-to-reproduce-our-training-recipes">How To Reproduce Our Training Recipes</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#professional-tools-integration">Professional Tools Integration</a></li>
<li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#supergradients-faq">SuperGradients FAQ</a></li>
</ul>

        </div>
      </div>
    </nav>

    <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
          <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
          <a href="../../../../index.html">SuperGradients</a>
      </nav>

      <div class="wy-nav-content">
        <div class="rst-content">
          <div role="navigation" aria-label="Page navigation">
  <ul class="wy-breadcrumbs">
      <li><a href="../../../../index.html" class="icon icon-home"></a> &raquo;</li>
          <li><a href="../../../index.html">Module code</a> &raquo;</li>
      <li>super_gradients.training.kd_model.kd_model</li>
      <li class="wy-breadcrumbs-aside">
      </li>
  </ul>
  <hr/>
</div>
          <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
           <div itemprop="articleBody">
             
  <h1>Source code for super_gradients.training.kd_model.kd_model</h1><div class="highlight"><pre>
<span></span><span class="kn">import</span> <span class="nn">torch.nn</span>

<span class="kn">from</span> <span class="nn">super_gradients.training.models.all_architectures</span> <span class="kn">import</span> <span class="n">KD_ARCHITECTURES</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.models.kd_modules.kd_module</span> <span class="kn">import</span> <span class="n">KDModule</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.sg_model</span> <span class="kn">import</span> <span class="n">SgModel</span>
<span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Union</span>
<span class="kn">from</span> <span class="nn">super_gradients.common.abstractions.abstract_logger</span> <span class="kn">import</span> <span class="n">get_logger</span>
<span class="kn">from</span> <span class="nn">super_gradients.training</span> <span class="kn">import</span> <span class="n">utils</span> <span class="k">as</span> <span class="n">core_utils</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.pretrained_models</span> <span class="kn">import</span> <span class="n">PRETRAINED_NUM_CLASSES</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.utils</span> <span class="kn">import</span> <span class="n">get_param</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.utils.checkpoint_utils</span> <span class="kn">import</span> <span class="n">read_ckpt_state_dict</span><span class="p">,</span> \
    <span class="n">load_checkpoint_to_model</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.exceptions.kd_model_exceptions</span> <span class="kn">import</span> <span class="n">ArchitectureKwargsException</span><span class="p">,</span> \
    <span class="n">UnsupportedKDArchitectureException</span><span class="p">,</span> <span class="n">InconsistentParamsException</span><span class="p">,</span> <span class="n">UnsupportedKDModelArgException</span><span class="p">,</span> \
    <span class="n">TeacherKnowledgeException</span><span class="p">,</span> <span class="n">UndefinedNumClassesException</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.utils.callbacks</span> <span class="kn">import</span> <span class="n">KDModelMetricsUpdateCallback</span>
<span class="kn">from</span> <span class="nn">super_gradients.training.utils.ema</span> <span class="kn">import</span> <span class="n">KDModelEMA</span>
<span class="n">logger</span> <span class="o">=</span> <span class="n">get_logger</span><span class="p">(</span><span class="vm">__name__</span><span class="p">)</span>


<div class="viewcode-block" id="KDModel"><a class="viewcode-back" href="../../../../super_gradients.training.html#super_gradients.training.KDModel">[docs]</a><span class="k">class</span> <span class="nc">KDModel</span><span class="p">(</span><span class="n">SgModel</span><span class="p">):</span>
    <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
        <span class="nb">super</span><span class="p">(</span><span class="n">KDModel</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">student_architecture</span> <span class="o">=</span> <span class="kc">None</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">teacher_architecture</span> <span class="o">=</span> <span class="kc">None</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">student_arch_params</span> <span class="o">=</span> <span class="kc">None</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">teacher_arch_params</span> <span class="o">=</span> <span class="kc">None</span>

<div class="viewcode-block" id="KDModel.build_model"><a class="viewcode-back" href="../../../../super_gradients.training.html#super_gradients.training.KDModel.build_model">[docs]</a>    <span class="k">def</span> <span class="nf">build_model</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span>
                    <span class="c1"># noqa: C901 - too complex</span>
                    <span class="n">architecture</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="nb">str</span><span class="p">,</span> <span class="n">KDModule</span><span class="p">]</span> <span class="o">=</span> <span class="s1">&#39;kd_module&#39;</span><span class="p">,</span>
                    <span class="n">arch_params</span><span class="o">=</span><span class="p">{},</span> <span class="n">checkpoint_params</span><span class="o">=</span><span class="p">{},</span>
                    <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        :param architecture: (Union[str, KDModule]) Defines the network&#39;s architecture from models/KD_ARCHITECTURES</span>
<span class="sd">         (default=&#39;kd_module&#39;)</span>

<span class="sd">        :param arch_params: (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc to be passed to kd</span>
<span class="sd">            architecture class (discarded when architecture is KDModule instance)</span>

<span class="sd">        :param checkpoint_params: (dict) A dictionary like object with the following keys/values:</span>

<span class="sd">              student_pretrained_weights:   String describing the dataset of the pretrained weights (for example</span>
<span class="sd">              &quot;imagenent&quot;) for the student network.</span>

<span class="sd">              teacher_pretrained_weights:   String describing the dataset of the pretrained weights (for example</span>
<span class="sd">              &quot;imagenent&quot;) for the teacher network.</span>

<span class="sd">              teacher_checkpoint_path:    Local path to the teacher&#39;s checkpoint. Note that when passing pretrained_weights</span>
<span class="sd">                                   through teacher_arch_params these weights will be overridden by the</span>
<span class="sd">                                   pretrained checkpoint. (default=None)</span>

<span class="sd">              load_kd_model_checkpoint:   Whether to load an entire KDModule checkpoint (used to continue KD training)</span>
<span class="sd">               (default=False)</span>

<span class="sd">              kd_model_source_ckpt_folder_name: Folder name to load an entire KDModule checkpoint from</span>
<span class="sd">                (self.experiment_name if none is given) to resume KD training (default=None)</span>

<span class="sd">              kd_model_external_checkpoint_path: The path to the external checkpoint to be loaded. Can be absolute or relative</span>
<span class="sd">                                               (ie: path/to/checkpoint.pth). If provided, will automatically attempt to</span>
<span class="sd">                                               load the checkpoint even if the load_checkpoint flag is not provided.</span>
<span class="sd">                                               (deafult=None)</span>

<span class="sd">        :keyword student_architecture: (Union[str, SgModule]) Defines the student&#39;s architecture from</span>
<span class="sd">            models/ALL_ARCHITECTURES (when str), or directly defined the student network (when SgModule).</span>

<span class="sd">        :keyword teacher_architecture: (Union[str, SgModule]) Defines the teacher&#39;s architecture from</span>
<span class="sd">            models/ALL_ARCHITECTURES (when str), or directly defined the teacher network (when SgModule).</span>

<span class="sd">        :keyword student_arch_params: (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for student</span>
<span class="sd">            net. (deafult={})</span>

<span class="sd">        :keyword teacher_arch_params: (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for teacher</span>
<span class="sd">            net. (deafult={})</span>

<span class="sd">        :keyword run_teacher_on_eval: (bool)- whether to run self.teacher at eval mode regardless of self.train(mode)</span>


<span class="sd">        &quot;&quot;&quot;</span>
        <span class="n">kwargs</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s2">&quot;student_architecture&quot;</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
        <span class="n">kwargs</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s2">&quot;teacher_architecture&quot;</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
        <span class="n">kwargs</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s2">&quot;student_arch_params&quot;</span><span class="p">,</span> <span class="p">{})</span>
        <span class="n">kwargs</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">,</span> <span class="p">{})</span>
        <span class="n">kwargs</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s2">&quot;run_teacher_on_eval&quot;</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">_validate_args</span><span class="p">(</span><span class="n">arch_params</span><span class="p">,</span> <span class="n">architecture</span><span class="p">,</span> <span class="n">checkpoint_params</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>

        <span class="bp">self</span><span class="o">.</span><span class="n">student_architecture</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;student_architecture&quot;</span><span class="p">)</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">teacher_architecture</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;teacher_architecture&quot;</span><span class="p">)</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">student_arch_params</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;student_arch_params&quot;</span><span class="p">)</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">teacher_arch_params</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">)</span>

        <span class="nb">super</span><span class="p">(</span><span class="n">KDModel</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">build_model</span><span class="p">(</span><span class="n">architecture</span><span class="o">=</span><span class="n">architecture</span><span class="p">,</span> <span class="n">arch_params</span><span class="o">=</span><span class="n">arch_params</span><span class="p">,</span>
                                         <span class="n">checkpoint_params</span><span class="o">=</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span></div>

    <span class="k">def</span> <span class="nf">_validate_args</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">arch_params</span><span class="p">,</span> <span class="n">architecture</span><span class="p">,</span> <span class="n">checkpoint_params</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
        <span class="n">student_architecture</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;student_architecture&quot;</span><span class="p">)</span>
        <span class="n">teacher_architecture</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;teacher_architecture&quot;</span><span class="p">)</span>
        <span class="n">student_arch_params</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;student_arch_params&quot;</span><span class="p">)</span>
        <span class="n">teacher_arch_params</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">)</span>

        <span class="k">if</span> <span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s1">&#39;pretrained_weights&#39;</span><span class="p">)</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
            <span class="k">raise</span> <span class="n">UnsupportedKDModelArgException</span><span class="p">(</span><span class="s2">&quot;pretrained_weights&quot;</span><span class="p">,</span> <span class="s2">&quot;checkpoint_params&quot;</span><span class="p">)</span>

        <span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">architecture</span><span class="p">,</span> <span class="n">KDModule</span><span class="p">):</span>
            <span class="k">if</span> <span class="n">student_architecture</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="n">teacher_architecture</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
                <span class="k">raise</span> <span class="n">ArchitectureKwargsException</span><span class="p">()</span>
            <span class="k">if</span> <span class="n">architecture</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">KD_ARCHITECTURES</span><span class="o">.</span><span class="n">keys</span><span class="p">():</span>
                <span class="k">raise</span> <span class="n">UnsupportedKDArchitectureException</span><span class="p">(</span><span class="n">architecture</span><span class="p">)</span>

        <span class="c1"># DERIVE NUMBER OF CLASSES FROM DATASET INTERFACE IF NOT SPECIFIED OR ARCH PARAMS FOR TEACHER AND STUDENT</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">_validate_num_classes</span><span class="p">(</span><span class="n">student_arch_params</span><span class="p">,</span> <span class="n">teacher_arch_params</span><span class="p">)</span>

        <span class="n">arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]</span> <span class="o">=</span> <span class="n">student_arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]</span>

        <span class="c1"># MAKE SURE TEACHER&#39;S PRETRAINED NUM CLASSES EQUALS TO THE ONES BELONGING TO STUDENT AS WE CAN&#39;T REPLACE</span>
        <span class="c1"># THE TEACHER&#39;S HEAD</span>
        <span class="n">teacher_pretrained_weights</span> <span class="o">=</span> <span class="n">core_utils</span><span class="o">.</span><span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s1">&#39;teacher_pretrained_weights&#39;</span><span class="p">,</span>
                                                          <span class="n">default_val</span><span class="o">=</span><span class="kc">None</span><span class="p">)</span>
        <span class="k">if</span> <span class="n">teacher_pretrained_weights</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
            <span class="n">teacher_pretrained_num_classes</span> <span class="o">=</span> <span class="n">PRETRAINED_NUM_CLASSES</span><span class="p">[</span><span class="n">teacher_pretrained_weights</span><span class="p">]</span>
            <span class="k">if</span> <span class="n">teacher_pretrained_num_classes</span> <span class="o">!=</span> <span class="n">teacher_arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]:</span>
                <span class="k">raise</span> <span class="n">InconsistentParamsException</span><span class="p">(</span><span class="s2">&quot;Pretrained dataset number of classes&quot;</span><span class="p">,</span> <span class="s2">&quot;teacher&#39;s arch params&quot;</span><span class="p">,</span>
                                                  <span class="s2">&quot;number of classes&quot;</span><span class="p">,</span> <span class="s2">&quot;student&#39;s number of classes&quot;</span><span class="p">)</span>

        <span class="n">teacher_checkpoint_path</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s2">&quot;teacher_checkpoint_path&quot;</span><span class="p">)</span>
        <span class="n">load_kd_model_checkpoint</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s2">&quot;load_checkpoint&quot;</span><span class="p">)</span>

        <span class="c1"># CHECK THAT TEACHER NETWORK HOLDS KNOWLEDGE FOR THE STUDENT TO LEARN FROM OR THAT WE ARE LOADING AN ENTIRE KD</span>
        <span class="k">if</span> <span class="ow">not</span> <span class="p">(</span><span class="n">teacher_pretrained_weights</span> <span class="ow">or</span> <span class="n">teacher_checkpoint_path</span> <span class="ow">or</span> <span class="n">load_kd_model_checkpoint</span> <span class="ow">or</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">teacher_architecture</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">)):</span>
            <span class="k">raise</span> <span class="n">TeacherKnowledgeException</span><span class="p">()</span>

    <span class="k">def</span> <span class="nf">_validate_num_classes</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">student_arch_params</span><span class="p">,</span> <span class="n">teacher_arch_params</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Checks validity of num_classes for num_classes (i.e existence and consistency between subnets)</span>

<span class="sd">        :param student_arch_params: (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for student</span>
<span class="sd">        :param teacher_arch_params: (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for teacher</span>

<span class="sd">        &quot;&quot;&quot;</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">_validate_subnet_num_classes</span><span class="p">(</span><span class="n">student_arch_params</span><span class="p">)</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">_validate_subnet_num_classes</span><span class="p">(</span><span class="n">teacher_arch_params</span><span class="p">)</span>
        <span class="k">if</span> <span class="n">teacher_arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]</span> <span class="o">!=</span> <span class="n">student_arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]:</span>
            <span class="k">raise</span> <span class="n">InconsistentParamsException</span><span class="p">(</span><span class="s2">&quot;num_classes&quot;</span><span class="p">,</span> <span class="s2">&quot;student_arch_params&quot;</span><span class="p">,</span> <span class="s2">&quot;num_classes&quot;</span><span class="p">,</span>
                                              <span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">)</span>

    <span class="k">def</span> <span class="nf">_validate_subnet_num_classes</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">subnet_arch_params</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Derives num_classes in student_arch_params/teacher_arch_params from dataset interface or raises an error</span>
<span class="sd">         when none is given</span>

<span class="sd">        :param subnet_arch_params: Arch params for student/teacher</span>

<span class="sd">        &quot;&quot;&quot;</span>

        <span class="k">if</span> <span class="s1">&#39;num_classes&#39;</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">subnet_arch_params</span><span class="o">.</span><span class="n">keys</span><span class="p">():</span>
            <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">dataset_interface</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
                <span class="k">raise</span> <span class="n">UndefinedNumClassesException</span><span class="p">()</span>
            <span class="k">else</span><span class="p">:</span>
                <span class="n">subnet_arch_params</span><span class="p">[</span><span class="s1">&#39;num_classes&#39;</span><span class="p">]</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">classes</span><span class="p">)</span>

    <span class="k">def</span> <span class="nf">_instantiate_net</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">architecture</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="n">KDModule</span><span class="p">,</span> <span class="n">KDModule</span><span class="o">.</span><span class="vm">__class__</span><span class="p">,</span> <span class="nb">str</span><span class="p">],</span> <span class="n">arch_params</span><span class="p">:</span> <span class="nb">dict</span><span class="p">,</span>
                         <span class="n">checkpoint_params</span><span class="p">:</span> <span class="nb">dict</span><span class="p">,</span> <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">tuple</span><span class="p">:</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Instantiates kd_module according to architecture and arch_params, handles pretrained weights for the student</span>
<span class="sd">         and teacher networks, and the required module manipulation (i.e head replacement) for the teacher network.</span>

<span class="sd">        :param architecture: String, KDModule or uninstantiated KDModule class describing the netowrks architecture.</span>
<span class="sd">        :param arch_params: Architecture&#39;s parameters passed to networks c&#39;tor.</span>
<span class="sd">        :param checkpoint_params: checkpoint loading related parameters dictionary with &#39;pretrained_weights&#39; key,</span>
<span class="sd">            s.t it&#39;s value is a string describing the dataset of the pretrained weights (for example &quot;imagenent&quot;).</span>

<span class="sd">        :return: instantiated netowrk i.e KDModule, architecture_class (will be none when architecture is not str)</span>
<span class="sd">        &quot;&quot;&quot;</span>

        <span class="n">student_architecture</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;student_architecture&quot;</span><span class="p">)</span>
        <span class="n">teacher_architecture</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;teacher_architecture&quot;</span><span class="p">)</span>
        <span class="n">student_arch_params</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;student_arch_params&quot;</span><span class="p">)</span>
        <span class="n">teacher_arch_params</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">)</span>
        <span class="n">student_arch_params</span> <span class="o">=</span> <span class="n">core_utils</span><span class="o">.</span><span class="n">HpmStruct</span><span class="p">(</span><span class="o">**</span><span class="n">student_arch_params</span><span class="p">)</span>
        <span class="n">teacher_arch_params</span> <span class="o">=</span> <span class="n">core_utils</span><span class="o">.</span><span class="n">HpmStruct</span><span class="p">(</span><span class="o">**</span><span class="n">teacher_arch_params</span><span class="p">)</span>
        <span class="n">student_pretrained_weights</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s1">&#39;student_pretrained_weights&#39;</span><span class="p">)</span>
        <span class="n">teacher_pretrained_weights</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s1">&#39;teacher_pretrained_weights&#39;</span><span class="p">)</span>

        <span class="n">student</span> <span class="o">=</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="n">_instantiate_net</span><span class="p">(</span><span class="n">student_architecture</span><span class="p">,</span> <span class="n">student_arch_params</span><span class="p">,</span>
                                           <span class="p">{</span><span class="s2">&quot;pretrained_weights&quot;</span><span class="p">:</span> <span class="n">student_pretrained_weights</span><span class="p">})</span>
        <span class="n">teacher</span> <span class="o">=</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="n">_instantiate_net</span><span class="p">(</span><span class="n">teacher_architecture</span><span class="p">,</span> <span class="n">teacher_arch_params</span><span class="p">,</span>
                                           <span class="p">{</span><span class="s2">&quot;pretrained_weights&quot;</span><span class="p">:</span> <span class="n">teacher_pretrained_weights</span><span class="p">})</span>

        <span class="n">run_teacher_on_eval</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">kwargs</span><span class="p">,</span> <span class="s2">&quot;run_teacher_on_eval&quot;</span><span class="p">,</span> <span class="n">default_val</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>

        <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">architecture</span><span class="p">,</span> <span class="nb">str</span><span class="p">):</span>
            <span class="n">architecture_cls</span> <span class="o">=</span> <span class="n">KD_ARCHITECTURES</span><span class="p">[</span><span class="n">architecture</span><span class="p">]</span>
            <span class="n">net</span> <span class="o">=</span> <span class="n">architecture_cls</span><span class="p">(</span><span class="n">arch_params</span><span class="o">=</span><span class="n">arch_params</span><span class="p">,</span> <span class="n">student</span><span class="o">=</span><span class="n">student</span><span class="p">,</span> <span class="n">teacher</span><span class="o">=</span><span class="n">teacher</span><span class="p">,</span>
                                   <span class="n">run_teacher_on_eval</span><span class="o">=</span><span class="n">run_teacher_on_eval</span><span class="p">)</span>
        <span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">architecture</span><span class="p">,</span> <span class="n">KDModule</span><span class="o">.</span><span class="vm">__class__</span><span class="p">):</span>
            <span class="n">net</span> <span class="o">=</span> <span class="n">architecture</span><span class="p">(</span><span class="n">arch_params</span><span class="o">=</span><span class="n">arch_params</span><span class="p">,</span> <span class="n">student</span><span class="o">=</span><span class="n">student</span><span class="p">,</span> <span class="n">teacher</span><span class="o">=</span><span class="n">teacher</span><span class="p">,</span>
                               <span class="n">run_teacher_on_eval</span><span class="o">=</span><span class="n">run_teacher_on_eval</span><span class="p">)</span>
        <span class="k">else</span><span class="p">:</span>
            <span class="n">net</span> <span class="o">=</span> <span class="n">architecture</span>

        <span class="k">return</span> <span class="n">net</span>

    <span class="k">def</span> <span class="nf">_load_checkpoint_to_model</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Initializes teacher weights with teacher_checkpoint_path if needed, then handles checkpoint loading for</span>
<span class="sd">         the entire KD network following the same logic as in SgModel.</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="n">teacher_checkpoint_path</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s2">&quot;teacher_checkpoint_path&quot;</span><span class="p">)</span>
        <span class="n">teacher_net</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">net</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">teacher</span>

        <span class="k">if</span> <span class="n">teacher_checkpoint_path</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>

            <span class="c1">#  WARN THAT TEACHER_CKPT WILL OVERRIDE TEACHER&#39;S PRETRAINED WEIGHTS</span>
            <span class="n">teacher_pretrained_weights</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">checkpoint_params</span><span class="p">,</span> <span class="s2">&quot;teacher_pretrained_weights&quot;</span><span class="p">)</span>
            <span class="k">if</span> <span class="n">teacher_pretrained_weights</span><span class="p">:</span>
                <span class="n">logger</span><span class="o">.</span><span class="n">warning</span><span class="p">(</span>
                    <span class="n">teacher_checkpoint_path</span> <span class="o">+</span> <span class="s2">&quot; checkpoint is &quot;</span>
                                              <span class="s2">&quot;overriding &quot;</span> <span class="o">+</span> <span class="n">teacher_pretrained_weights</span> <span class="o">+</span> <span class="s2">&quot; for teacher model&quot;</span><span class="p">)</span>

            <span class="c1"># ALWAYS LOAD ITS EMA IF IT EXISTS</span>
            <span class="n">load_teachers_ema</span> <span class="o">=</span> <span class="s1">&#39;ema_net&#39;</span> <span class="ow">in</span> <span class="n">read_ckpt_state_dict</span><span class="p">(</span><span class="n">teacher_checkpoint_path</span><span class="p">)</span><span class="o">.</span><span class="n">keys</span><span class="p">()</span>
            <span class="n">load_checkpoint_to_model</span><span class="p">(</span><span class="n">ckpt_local_path</span><span class="o">=</span><span class="n">teacher_checkpoint_path</span><span class="p">,</span>
                                     <span class="n">load_backbone</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span>
                                     <span class="n">net</span><span class="o">=</span><span class="n">teacher_net</span><span class="p">,</span>
                                     <span class="n">strict</span><span class="o">=</span><span class="s1">&#39;no_key_matching&#39;</span><span class="p">,</span>
                                     <span class="n">load_weights_only</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
                                     <span class="n">load_ema_as_net</span><span class="o">=</span><span class="n">load_teachers_ema</span><span class="p">)</span>

        <span class="nb">super</span><span class="p">(</span><span class="n">KDModel</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="n">_load_checkpoint_to_model</span><span class="p">()</span>

    <span class="k">def</span> <span class="nf">_add_metrics_update_callback</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">phase</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Adds KDModelMetricsUpdateCallback to be fired at phase</span>

<span class="sd">        :param phase: Phase for the metrics callback to be fired at</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">phase_callbacks</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">KDModelMetricsUpdateCallback</span><span class="p">(</span><span class="n">phase</span><span class="p">))</span>

    <span class="k">def</span> <span class="nf">_get_hyper_param_config</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Creates a training hyper param config for logging with additional KD related hyper params.</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="n">hyper_param_config</span> <span class="o">=</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="n">_get_hyper_param_config</span><span class="p">()</span>
        <span class="n">hyper_param_config</span><span class="o">.</span><span class="n">update</span><span class="p">({</span><span class="s2">&quot;student_architecture&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">student_architecture</span><span class="p">,</span>
                                   <span class="s2">&quot;teacher_architecture&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">teacher_architecture</span><span class="p">,</span>
                                   <span class="s2">&quot;student_arch_params&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">student_arch_params</span><span class="p">,</span>
                                   <span class="s2">&quot;teacher_arch_params&quot;</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">teacher_arch_params</span>
                                   <span class="p">})</span>
        <span class="k">return</span> <span class="n">hyper_param_config</span>

    <span class="k">def</span> <span class="nf">_instantiate_ema_model</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">decay</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.9999</span><span class="p">,</span> <span class="n">beta</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mi">15</span><span class="p">,</span> <span class="n">exp_activation</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">KDModelEMA</span><span class="p">:</span>
        <span class="sd">&quot;&quot;&quot;Instantiate KD ema model for KDModule.</span>

<span class="sd">        If the model is of class KDModule, the instance will be adapted to work on knowledge distillation.</span>
<span class="sd">        :param decay:           the maximum decay value. as the training process advances, the decay will climb towards</span>
<span class="sd">                                this value until the EMA_t+1 = EMA_t * decay + TRAINING_MODEL * (1- decay)</span>
<span class="sd">        :param beta:            the exponent coefficient. The higher the beta, the sooner in the training the decay will</span>
<span class="sd">                                saturate to its final value. beta=15 is ~40% of the training process.</span>
<span class="sd">        :param exp_activation:</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="k">return</span> <span class="n">KDModelEMA</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">net</span><span class="p">,</span> <span class="n">decay</span><span class="p">,</span> <span class="n">beta</span><span class="p">,</span> <span class="n">exp_activation</span><span class="p">)</span>

    <span class="k">def</span> <span class="nf">_save_best_checkpoint</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">epoch</span><span class="p">,</span> <span class="n">state</span><span class="p">):</span>
        <span class="sd">&quot;&quot;&quot;</span>
<span class="sd">        Overrides parent best_ckpt saving to modify the state dict so that we only save the student.</span>
<span class="sd">        &quot;&quot;&quot;</span>
        <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">ema</span><span class="p">:</span>
            <span class="n">best_net</span> <span class="o">=</span> <span class="n">core_utils</span><span class="o">.</span><span class="n">WrappedModel</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">ema_model</span><span class="o">.</span><span class="n">ema</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">student</span><span class="p">)</span>
            <span class="n">state</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="s2">&quot;ema_net&quot;</span><span class="p">)</span>
        <span class="k">else</span><span class="p">:</span>
            <span class="n">best_net</span> <span class="o">=</span> <span class="n">core_utils</span><span class="o">.</span><span class="n">WrappedModel</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">net</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">student</span><span class="p">)</span>

        <span class="n">state</span><span class="p">[</span><span class="s2">&quot;net&quot;</span><span class="p">]</span> <span class="o">=</span> <span class="n">best_net</span><span class="o">.</span><span class="n">state_dict</span><span class="p">()</span>
        <span class="bp">self</span><span class="o">.</span><span class="n">sg_logger</span><span class="o">.</span><span class="n">add_checkpoint</span><span class="p">(</span><span class="n">tag</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">ckpt_best_name</span><span class="p">,</span> <span class="n">state_dict</span><span class="o">=</span><span class="n">state</span><span class="p">,</span> <span class="n">global_step</span><span class="o">=</span><span class="n">epoch</span><span class="p">)</span></div>
</pre></div>

           </div>
          </div>
          <footer>

  <hr/>

  <div role="contentinfo">
    <p>&#169; Copyright 2021, SuperGradients team.</p>
  </div>

  Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    provided by <a href="https://readthedocs.org">Read the Docs</a>.
   

</footer>
        </div>
      </div>
    </section>
  </div>
  <script>
      jQuery(function () {
          SphinxRtdTheme.Navigation.enable(true);
      });
  </script> 

</body>
</html>