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

	
394

	
395

	
396

	
397

	
398

	
399

	
400

	
401

	
402

	
403

	
404

	
405

	
406

	
407

	
408

	
409

	
410

	
411

	
412

	
413

	
414

	
415

	
416

	
417

	
418

	
419

	
420

	
421

	
422

	
423

	
424

	
425

	
426

	
427

	
428

	
429

	
430

	
431

	
432

	
433

	
434

	
435

	
436

	
437

	
438

	
439

	
440

	
441

	
442

	
443

	
444

	
445

	
446

	
447

	
448

	
449

	
450

	
451

	
452

	
453

	
454

	
455

	
456

	
457

	
458

	
459

	
460

	
461

	
462

	
463

	
464

	
465

	
466

	
467

	
468

	
469

	
470

	
471

	
472

	
473

	
474

	
475

	
476

	
477

	
478

	
479

	
480

	
481

	
482

	
483

	
484

	
485

	
486

	
487

	
488

	
489

	
490

	
491

	
492

	
493

	
494

	
495

	
496

	
497

	
498

	
499

	
500

	
501

	
502

	
503

	
504

	
505

	
506

	
507

	
508

	
509

	
510

	
511

	
512

	
513

	
514

	
515

	
516

	
517

	
518

	
519

	
520

	
521

	
522

	
523

	
524

	
525

	
526

	
527

	
528

	
529

	
530

	
531

	
532

	
533

	
534

	
535

	
536

	
537

	
538

	
539

	
540

	
541

	
542

	
543

	
544

	
545

	
546

	
547

	
548

	
549

	
550

	
            <!DOCTYPE html>
<html class="writer-html5" lang="en" >
<head>
  <meta charset="utf-8" /><meta name="generator" content="Docutils 0.17.1: http://docutils.sourceforge.net/" />

  <meta name="viewport" content="width=device-width, initial-scale=1.0" />
  <title>super_gradients.training.losses package &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" />
  <!--[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" />
    <link rel="next" title="super_gradients.training.metrics package" href="super_gradients.training.metrics.html" />
    <link rel="prev" title="super_gradients.training.legacy package" href="super_gradients.training.legacy.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">Contents:</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="intro.html">Introduction</a></li>
<li class="toctree-l1"><a class="reference internal" href="intro.html#installation">Installation</a></li>
<li class="toctree-l1 current"><a class="reference internal" href="super_gradients.training.html">super_gradients.training package</a><ul class="current">
<li class="toctree-l2 current"><a class="reference internal" href="super_gradients.training.html#subpackages">Subpackages</a><ul class="current">
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.datasets.html">super_gradients.training.datasets package</a></li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.exceptions.html">super_gradients.training.exceptions package</a></li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.legacy.html">super_gradients.training.legacy package</a></li>
<li class="toctree-l3 current"><a class="current reference internal" href="#">super_gradients.training.losses package</a><ul>
<li class="toctree-l4"><a class="reference internal" href="#submodules">Submodules</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.all_losses">super_gradients.training.losses.all_losses module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.ddrnet_loss">super_gradients.training.losses.ddrnet_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.focal_loss">super_gradients.training.losses.focal_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.label_smoothing_cross_entropy_loss">super_gradients.training.losses.label_smoothing_cross_entropy_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.ohem_ce_loss">super_gradients.training.losses.ohem_ce_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.r_squared_loss">super_gradients.training.losses.r_squared_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.shelfnet_ohem_loss">super_gradients.training.losses.shelfnet_ohem_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.shelfnet_semantic_encoding_loss">super_gradients.training.losses.shelfnet_semantic_encoding_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.ssd_loss">super_gradients.training.losses.ssd_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.yolo_v3_loss">super_gradients.training.losses.yolo_v3_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses.yolo_v5_loss">super_gradients.training.losses.yolo_v5_loss module</a></li>
<li class="toctree-l4"><a class="reference internal" href="#module-super_gradients.training.losses">Module contents</a></li>
</ul>
</li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.metrics.html">super_gradients.training.metrics package</a></li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.models.html">super_gradients.training.models package</a></li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.sg_model.html">super_gradients.training.sg_model package</a></li>
<li class="toctree-l3"><a class="reference internal" href="super_gradients.training.utils.html">super_gradients.training.utils package</a></li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="super_gradients.training.html#submodules">Submodules</a></li>
<li class="toctree-l2"><a class="reference internal" href="super_gradients.training.html#module-super_gradients.training.params">super_gradients.training.params module</a></li>
<li class="toctree-l2"><a class="reference internal" href="super_gradients.training.html#module-super_gradients.training.pretrained_models">super_gradients.training.pretrained_models module</a></li>
<li class="toctree-l2"><a class="reference internal" href="super_gradients.training.html#module-super_gradients.training">Module contents</a></li>
</ul>
</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="super_gradients.training.html">super_gradients.training package</a> &raquo;</li>
      <li>super_gradients.training.losses package</li>
      <li class="wy-breadcrumbs-aside">
            <a href="_sources/super_gradients.training.losses.rst.txt" rel="nofollow"> View page source</a>
      </li>
  </ul>
  <hr/>
</div>
          <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
           <div itemprop="articleBody">
             
  <section id="super-gradients-training-losses-package">
<h1>super_gradients.training.losses package<a class="headerlink" href="#super-gradients-training-losses-package" title="Permalink to this headline"></a></h1>
<section id="submodules">
<h2>Submodules<a class="headerlink" href="#submodules" title="Permalink to this headline"></a></h2>
</section>
<section id="module-super_gradients.training.losses.all_losses">
<span id="super-gradients-training-losses-all-losses-module"></span><h2>super_gradients.training.losses.all_losses module<a class="headerlink" href="#module-super_gradients.training.losses.all_losses" title="Permalink to this headline"></a></h2>
</section>
<section id="module-super_gradients.training.losses.ddrnet_loss">
<span id="super-gradients-training-losses-ddrnet-loss-module"></span><h2>super_gradients.training.losses.ddrnet_loss module<a class="headerlink" href="#module-super_gradients.training.losses.ddrnet_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.ddrnet_loss.DDRNetLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.ddrnet_loss.</span></span><span class="sig-name descname"><span class="pre">DDRNetLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.7</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ohem_percentage</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weights</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">[1.0,</span> <span class="pre">0.4]</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_label</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">255</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ddrnet_loss.html#DDRNetLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ddrnet_loss.DDRNetLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <a class="reference internal" href="#super_gradients.training.losses.ohem_ce_loss.OhemCELoss" title="super_gradients.training.losses.ohem_ce_loss.OhemCELoss"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.losses.ohem_ce_loss.OhemCELoss</span></code></a></p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.ddrnet_loss.DDRNetLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions_list</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ddrnet_loss.html#DDRNetLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ddrnet_loss.DDRNetLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.ddrnet_loss.DDRNetLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.ddrnet_loss.DDRNetLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.focal_loss">
<span id="super-gradients-training-losses-focal-loss-module"></span><h2>super_gradients.training.losses.focal_loss module<a class="headerlink" href="#module-super_gradients.training.losses.focal_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.focal_loss.FocalLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.focal_loss.</span></span><span class="sig-name descname"><span class="pre">FocalLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">loss_fcn</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.loss.BCEWithLogitsLoss</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gamma</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.25</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/focal_loss.html#FocalLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.focal_loss.FocalLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<p>Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.focal_loss.FocalLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">pred</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">true</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/focal_loss.html#FocalLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.focal_loss.FocalLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.focal_loss.FocalLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.focal_loss.FocalLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.label_smoothing_cross_entropy_loss">
<span id="super-gradients-training-losses-label-smoothing-cross-entropy-loss-module"></span><h2>super_gradients.training.losses.label_smoothing_cross_entropy_loss module<a class="headerlink" href="#module-super_gradients.training.losses.label_smoothing_cross_entropy_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.label_smoothing_cross_entropy_loss.</span></span><span class="sig-name descname"><span class="pre">LabelSmoothingCrossEntropyLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'mean'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_dist</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">from_logits</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#LabelSmoothingCrossEntropyLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss.CrossEntropyLoss</span></code></p>
<p>CrossEntropyLoss - with ability to recieve distrbution as targets, and optional label smoothing</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_dist</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#LabelSmoothingCrossEntropyLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss.ignore_index">
<span class="sig-name descname"><span class="pre">ignore_index</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.losses.label_smoothing_cross_entropy_loss.LabelSmoothingCrossEntropyLoss.ignore_index" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

<dl class="py function">
<dt class="sig sig-object py" id="super_gradients.training.losses.label_smoothing_cross_entropy_loss.cross_entropy">
<span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.label_smoothing_cross_entropy_loss.</span></span><span class="sig-name descname"><span class="pre">cross_entropy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">inputs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'mean'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_dist</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">from_logits</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#cross_entropy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.label_smoothing_cross_entropy_loss.cross_entropy" title="Permalink to this definition"></a></dt>
<dd><p>cross entropy loss, with support for target distributions and label smoothing <a class="reference external" href="https://arxiv.org/abs/1512.00567">https://arxiv.org/abs/1512.00567</a></p>
</dd></dl>

<dl class="py function">
<dt class="sig sig-object py" id="super_gradients.training.losses.label_smoothing_cross_entropy_loss.onehot">
<span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.label_smoothing_cross_entropy_loss.</span></span><span class="sig-name descname"><span class="pre">onehot</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">indexes</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">N</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#onehot"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.label_smoothing_cross_entropy_loss.onehot" title="Permalink to this definition"></a></dt>
<dd><p>Creates a one-hot representation of indexes with N possible entries
if N is not specified, it will suit the maximum index appearing.
indexes is a long-tensor of indexes
ignore_index will be zero in onehot representation</p>
</dd></dl>

</section>
<section id="module-super_gradients.training.losses.ohem_ce_loss">
<span id="super-gradients-training-losses-ohem-ce-loss-module"></span><h2>super_gradients.training.losses.ohem_ce_loss module<a class="headerlink" href="#module-super_gradients.training.losses.ohem_ce_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.ohem_ce_loss.OhemCELoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.ohem_ce_loss.</span></span><span class="sig-name descname"><span class="pre">OhemCELoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mining_percent</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_lb</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_pixels_exclude_ignored</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ohem_ce_loss.html#OhemCELoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ohem_ce_loss.OhemCELoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<p>OhemCELoss - Online Hard Example Mining Cross Entropy Loss</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.ohem_ce_loss.OhemCELoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">logits</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">labels</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ohem_ce_loss.html#OhemCELoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ohem_ce_loss.OhemCELoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.ohem_ce_loss.OhemCELoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.ohem_ce_loss.OhemCELoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.r_squared_loss">
<span id="super-gradients-training-losses-r-squared-loss-module"></span><h2>super_gradients.training.losses.r_squared_loss module<a class="headerlink" href="#module-super_gradients.training.losses.r_squared_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.r_squared_loss.RSquaredLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.r_squared_loss.</span></span><span class="sig-name descname"><span class="pre">RSquaredLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">size_average</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduce</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'mean'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/r_squared_loss.html#RSquaredLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.r_squared_loss.RSquaredLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.r_squared_loss.RSquaredLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/r_squared_loss.html#RSquaredLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.r_squared_loss.RSquaredLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Computes the R-squared for the output and target values
:param output: Tensor / Numpy / List</p>
<blockquote>
<div><p>The prediction</p>
</div></blockquote>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>target</strong> – Tensor / Numpy / List
The corresponding lables</p>
</dd>
</dl>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.r_squared_loss.RSquaredLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.r_squared_loss.RSquaredLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.shelfnet_ohem_loss">
<span id="super-gradients-training-losses-shelfnet-ohem-loss-module"></span><h2>super_gradients.training.losses.shelfnet_ohem_loss module<a class="headerlink" href="#module-super_gradients.training.losses.shelfnet_ohem_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.shelfnet_ohem_loss.</span></span><span class="sig-name descname"><span class="pre">ShelfNetOHEMLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.7</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mining_percent</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.0001</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_lb</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">255</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_ohem_loss.html#ShelfNetOHEMLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <a class="reference internal" href="#super_gradients.training.losses.ohem_ce_loss.OhemCELoss" title="super_gradients.training.losses.ohem_ce_loss.OhemCELoss"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.losses.ohem_ce_loss.OhemCELoss</span></code></a></p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions_list</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_ohem_loss.html#ShelfNetOHEMLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.training">
<span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.losses.shelfnet_ohem_loss.ShelfNetOHEMLoss.training" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.shelfnet_semantic_encoding_loss">
<span id="super-gradients-training-losses-shelfnet-semantic-encoding-loss-module"></span><h2>super_gradients.training.losses.shelfnet_semantic_encoding_loss module<a class="headerlink" href="#module-super_gradients.training.losses.shelfnet_semantic_encoding_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.shelfnet_semantic_encoding_loss.</span></span><span class="sig-name descname"><span class="pre">ShelfNetSemanticEncodingLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">se_weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">nclass</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">21</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">aux_weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.4</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">1</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_semantic_encoding_loss.html#ShelfNetSemanticEncodingLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss.CrossEntropyLoss</span></code></p>
<p>2D Cross Entropy Loss with Auxilary Loss</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">logits</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">labels</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_semantic_encoding_loss.html#ShelfNetSemanticEncodingLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss.ignore_index">
<span class="sig-name descname"><span class="pre">ignore_index</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.losses.shelfnet_semantic_encoding_loss.ShelfNetSemanticEncodingLoss.ignore_index" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.ssd_loss">
<span id="super-gradients-training-losses-ssd-loss-module"></span><h2>super_gradients.training.losses.ssd_loss module<a class="headerlink" href="#module-super_gradients.training.losses.ssd_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.ssd_loss.SSDLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.ssd_loss.</span></span><span class="sig-name descname"><span class="pre">SSDLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dboxes</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.utils.html#super_gradients.training.utils.ssd_utils.DefaultBoxes" title="super_gradients.training.utils.ssd_utils.DefaultBoxes"><span class="pre">super_gradients.training.utils.ssd_utils.DefaultBoxes</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ssd_loss.SSDLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<p>Implements the loss as the sum of the followings:
1. Confidence Loss: All labels, with hard negative mining
2. Localization Loss: Only on positive labels</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.ssd_loss.SSDLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ssd_loss.SSDLoss.forward" title="Permalink to this definition"></a></dt>
<dd><dl class="simple">
<dt>Compute the loss</dt><dd><p>:param predictions - predictions tensor coming from the network. shape [N, num_classes+4, num_dboxes]
were the first four items are (x,y,w,h) and the rest are class confidence
:param targets - targets for the batch. [num targets, 6] (index in batch, label, x,y,w,h)</p>
</dd>
</dl>
</dd></dl>

<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.ssd_loss.SSDLoss.match_dboxes">
<span class="sig-name descname"><span class="pre">match_dboxes</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss.match_dboxes"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ssd_loss.SSDLoss.match_dboxes" title="Permalink to this definition"></a></dt>
<dd><p>convert ground truth boxes into a tensor with the same size as dboxes. each gt bbox is matched to every
destination box which overlaps it over 0.5 (IoU). so some gt bboxes can be duplicated to a few destination boxes
:param targets: a tensor containing the boxes for a single image. shape [num_boxes, 5] (x,y,w,h,label)
:return: two tensors</p>
<blockquote>
<div><p>boxes - shape of dboxes [4, num_dboxes] (x,y,w,h)
labels - sahpe [num_dboxes]</p>
</div></blockquote>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.ssd_loss.SSDLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.ssd_loss.SSDLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.yolo_v3_loss">
<span id="super-gradients-training-losses-yolo-v3-loss-module"></span><h2>super_gradients.training.losses.yolo_v3_loss module<a class="headerlink" href="#module-super_gradients.training.losses.yolo_v3_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.yolo_v3_loss.</span></span><span class="sig-name descname"><span class="pre">YoLoV3DetectionLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls_pw</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">obj_pw</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">giou</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">3.54</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">obj</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">64.3</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">37.4</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v3_loss.html#YoLoV3DetectionLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<p>YoLoV3DetectionLoss - Loss Class for Object Detection</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v3_loss.html#YoLoV3DetectionLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.yolo_v3_loss.YoLoV3DetectionLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses.yolo_v5_loss">
<span id="super-gradients-training-losses-yolo-v5-loss-module"></span><h2>super_gradients.training.losses.yolo_v5_loss module<a class="headerlink" href="#module-super_gradients.training.losses.yolo_v5_loss" title="Permalink to this headline"></a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss">
<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.yolo_v5_loss.</span></span><span class="sig-name descname"><span class="pre">YoLoV5DetectionLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">anchors</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.Anchors" title="super_gradients.training.utils.detection_utils.Anchors"><span class="pre">super_gradients.training.utils.detection_utils.Anchors</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls_pos_weight</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">obj_pos_weight</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">obj_loss_gain</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">box_loss_gain</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.05</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls_loss_gain</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">focal_loss_gamma</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls_objectness_weights</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v5_loss.html#YoLoV5DetectionLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss" title="Permalink to this definition"></a></dt>
<dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
<p>Calculate YOLO V5 loss:
L = L_objectivness + L_boxes + L_classification</p>
<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.build_targets">
<span class="sig-name descname"><span class="pre">build_targets</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">anchor_threshold</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">4.0</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v5_loss.html#YoLoV5DetectionLoss.build_targets"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.build_targets" title="Permalink to this definition"></a></dt>
<dd><dl>
<dt>Assign targets to anchors to use in L_boxes &amp; L_classification calculation:</dt><dd><ul class="simple">
<li><p>each target can be assigned to a few anchors,</p></li>
</ul>
<p>all anchors that are within [1/anchor_threshold, anchor_threshold] times target size range
* each anchor can be assigned to a few targets</p>
</dd>
</dl>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>predictions</strong> – Yolo predictions</p></li>
<li><p><strong>targets</strong> – ground truth targets</p></li>
<li><p><strong>anchor_threshold</strong> – ratio defining a size range of an appropriate anchor</p></li>
</ul>
</dd>
<dt class="field-even">Returns</dt>
<dd class="field-even"><p><p>each of 4 outputs contains one element for each Yolo output,
correspondences are raveled over the whole batch and all anchors:</p>
<blockquote>
<div><ul class="simple">
<li><p>classes of the targets;</p></li>
<li><p>boxes of the targets;</p></li>
<li><p>image id in a batch, anchor id, grid y, grid x coordinates;</p></li>
<li><p>anchor sizes.</p></li>
</ul>
</div></blockquote>
<p>All the above can be indexed in parallel to get the selected correspondences</p>
</p>
</dd>
</dl>
</dd></dl>

<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.compute_loss">
<span class="sig-name descname"><span class="pre">compute_loss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">giou_loss_ratio</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v5_loss.html#YoLoV5DetectionLoss.compute_loss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.compute_loss" title="Permalink to this definition"></a></dt>
<dd><p>L = L_objectivness + L_boxes + L_classification
where:</p>
<blockquote>
<div><ul class="simple">
<li><p>L_boxes and L_classification are calculated only between anchors and targets that suit them;</p></li>
<li><p>L_objectivness is calculated on all anchors.</p></li>
</ul>
</div></blockquote>
<dl>
<dt>L_classification:</dt><dd><p>for anchors that have suitable ground truths in their grid locations add BCEs
to force max probability for each GT class in a multi-label way
Coef: self.cls_loss_gain</p>
</dd>
<dt>L_boxes:</dt><dd><p>for anchors that have suitable ground truths in their grid locations
add (1 - IoU), IoU between a predicted box and each GT box, force maximum IoU
Coef: self.box_loss_gain</p>
</dd>
<dt>L_objectness:</dt><dd><p>for each anchor add BCE to force a prediction of (1 - giou_loss_ratio) + giou_loss_ratio * IoU,
IoU between a predicted box and random GT in it
Coef: self.obj_loss_gain, loss from each YOLO grid is additionally multiplied by balance = [4.0, 1.0, 0.4]</p>
<blockquote>
<div><p>to balance different contributions coming from different numbers of grid cells</p>
</div></blockquote>
</dd>
</dl>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>predictions</strong> – output from all Yolo levels, each of shape
[Batch x Num_Anchors x GridSizeY x GridSizeX x (4 + 1 + Num_classes)]</p></li>
<li><p><strong>targets</strong> – [Num_targets x (4 + 2)], values on dim 1 are: image id in a batch, class, box x y w h</p></li>
<li><p><strong>giou_loss_ratio</strong> – a coef in L_objectness defining what should be predicted as objecness
in a call with a target: can be a value in [IoU, 1] range</p></li>
</ul>
</dd>
<dt class="field-even">Returns</dt>
<dd class="field-even"><p>loss, all losses separately in a detached tensor</p>
</dd>
</dl>
</dd></dl>

<dl class="py method">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.forward">
<span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolo_v5_loss.html#YoLoV5DetectionLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.forward" title="Permalink to this definition"></a></dt>
<dd><p>Defines the computation performed at every call.</p>
<p>Should be overridden by all subclasses.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Although the recipe for forward pass needs to be defined within
this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
instead of this since the former takes care of running the
registered hooks while the latter silently ignores them.</p>
</div>
</dd></dl>

<dl class="py attribute">
<dt class="sig sig-object py" id="super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.reduction">
<span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.yolo_v5_loss.YoLoV5DetectionLoss.reduction" title="Permalink to this definition"></a></dt>
<dd></dd></dl>

</dd></dl>

</section>
<section id="module-super_gradients.training.losses">
<span id="module-contents"></span><h2>Module contents<a class="headerlink" href="#module-super_gradients.training.losses" title="Permalink to this headline"></a></h2>
</section>
</section>


           </div>
          </div>
          <footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
        <a href="super_gradients.training.legacy.html" class="btn btn-neutral float-left" title="super_gradients.training.legacy package" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
        <a href="super_gradients.training.metrics.html" class="btn btn-neutral float-right" title="super_gradients.training.metrics package" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
    </div>

  <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>