Register
Login
Resources
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Product
Data Engine LLMs Platform Enterprise
Pricing Explore
Connect to our Discord channel
Resources  
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Explore Pricing
Product  
Data Engine LLMs Platform Enterprise
Connect to Dagshub Discord channel
Register
Login

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

Watch 4
Issues 62 Pull Requests 44 Discussions 0

yolov5.html 103 KB
History Raw

You have to be logged in to leave a comment. Sign In
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 
  1. <!DOCTYPE html>
    2. 

  2. <html class="writer-html5" lang="en" >
    3. 

  3. <head>
    4. 

  4.   <meta charset="utf-8" />
    5. 

  5.   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
    6. 

  6.   <title>super_gradients.training.models.yolov5 &mdash; SuperGradients 1.0 documentation</title>
    7. 

  7.       <link rel="stylesheet" href="../../../../_static/pygments.css" type="text/css" />
    8. 

  8.       <link rel="stylesheet" href="../../../../_static/css/theme.css" type="text/css" />
    9. 

  9.   <!--[if lt IE 9]>
    10. 

  10.     <script src="../../../../_static/js/html5shiv.min.js"></script>
    11. 

  11.   <![endif]-->
    12. 

  12.   
    13. 

  13.         <script data-url_root="../../../../" id="documentation_options" src="../../../../_static/documentation_options.js"></script>
    14. 

  14.         <script src="../../../../_static/jquery.js"></script>
    15. 

  15.         <script src="../../../../_static/underscore.js"></script>
    16. 

  16.         <script src="../../../../_static/doctools.js"></script>
    17. 

  17.     <script src="../../../../_static/js/theme.js"></script>
    18. 

  18.     <link rel="index" title="Index" href="../../../../genindex.html" />
    19. 

  19.     <link rel="search" title="Search" href="../../../../search.html" /> 
    20. 

  20. </head>
    21. 

  21. 

  22. <body class="wy-body-for-nav"> 
    23. 

  23.   <div class="wy-grid-for-nav">
    24. 

  24.     <nav data-toggle="wy-nav-shift" class="wy-nav-side">
    25. 

  25.       <div class="wy-side-scroll">
    26. 

  26.         <div class="wy-side-nav-search" >
    27. 

  27.             <a href="../../../../index.html" class="icon icon-home"> SuperGradients
    28. 

  28.           </a>
    29. 

  29. <div role="search">
    30. 

  30.   <form id="rtd-search-form" class="wy-form" action="../../../../search.html" method="get">
    31. 

  31.     <input type="text" name="q" placeholder="Search docs" />
    32. 

  32.     <input type="hidden" name="check_keywords" value="yes" />
    33. 

  33.     <input type="hidden" name="area" value="default" />
    34. 

  34.   </form>
    35. 

  35. </div>
    36. 

  36.         </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
    37. 

  37.               <p class="caption"><span class="caption-text">Contents:</span></p>
    38. 

  38. <ul>
    39. 

  39. <li class="toctree-l1"><a class="reference internal" href="../../../../intro.html">Introduction</a></li>
    40. 

  40. <li class="toctree-l1"><a class="reference internal" href="../../../../intro.html#installation">Installation</a></li>
    41. 

  41. <li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.training.html">super_gradients.training package</a></li>
    42. 

  42. </ul>
    43. 

  43. 

  44.         </div>
    45. 

  45.       </div>
    46. 

  46.     </nav>
    47. 

  47. 

  48.     <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
    49. 

  49.           <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
    50. 

  50.           <a href="../../../../index.html">SuperGradients</a>
    51. 

  51.       </nav>
    52. 

  52. 

  53.       <div class="wy-nav-content">
    54. 

  54.         <div class="rst-content">
    55. 

  55.           <div role="navigation" aria-label="Page navigation">
    56. 

  56.   <ul class="wy-breadcrumbs">
    57. 

  57.       <li><a href="../../../../index.html" class="icon icon-home"></a> &raquo;</li>
    58. 

  58.           <li><a href="../../../index.html">Module code</a> &raquo;</li>
    59. 

  59.       <li>super_gradients.training.models.yolov5</li>
    60. 

  60.       <li class="wy-breadcrumbs-aside">
    61. 

  61.       </li>
    62. 

  62.   </ul>
    63. 

  63.   <hr/>
    64. 

  64. </div>
    65. 

  65.           <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
    66. 

  66.            <div itemprop="articleBody">
    67. 

  67.              
    68. 

  68.   <h1>Source code for super_gradients.training.models.yolov5</h1><div class="highlight"><pre>
    69. 

  69. <span></span><span class="sd">&quot;&quot;&quot;</span>
    70. 

  70. <span class="sd">YoloV5 code adapted from https://github.com/ultralytics/yolov5/blob/master/models/yolo.py</span>
    71. 

  71. <span class="sd">&quot;&quot;&quot;</span>
    72. 

  72. <span class="kn">import</span> <span class="nn">math</span>
    73. 

  73. <span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Union</span><span class="p">,</span> <span class="n">Type</span><span class="p">,</span> <span class="n">List</span>
    74. 

  74. 

  75. <span class="kn">import</span> <span class="nn">torch</span>
    76. 

  76. <span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
    77. 

  77. <span class="kn">from</span> <span class="nn">super_gradients.training.models.csp_darknet53</span> <span class="kn">import</span> <span class="n">width_multiplier</span><span class="p">,</span> <span class="n">Conv</span><span class="p">,</span> <span class="n">BottleneckCSP</span><span class="p">,</span> <span class="n">CSPDarknet53</span>
    78. 

  78. <span class="kn">from</span> <span class="nn">super_gradients.training.models.sg_module</span> <span class="kn">import</span> <span class="n">SgModule</span>
    79. 

  79. <span class="kn">from</span> <span class="nn">super_gradients.training.utils.detection_utils</span> <span class="kn">import</span> <span class="n">non_max_suppression</span><span class="p">,</span> <span class="n">scale_img</span><span class="p">,</span> \
    80. 

  80.     <span class="n">check_anchor_order</span><span class="p">,</span> <span class="n">check_img_size_divisibilty</span><span class="p">,</span> <span class="n">matrix_non_max_suppression</span><span class="p">,</span> <span class="n">NMS_Type</span><span class="p">,</span> \
    81. 

  81.     <span class="n">DetectionPostPredictionCallback</span><span class="p">,</span> <span class="n">Anchors</span>
    82. 

  82. <span class="kn">from</span> <span class="nn">super_gradients.training.utils.export_utils</span> <span class="kn">import</span> <span class="n">ExportableHardswish</span>
    83. 

  83. <span class="kn">from</span> <span class="nn">super_gradients.training.utils.utils</span> <span class="kn">import</span> <span class="n">HpmStruct</span><span class="p">,</span> <span class="n">get_param</span><span class="p">,</span> <span class="n">print_once</span>
    84. 

  84. <span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
    85. 

  85. 

  86. <span class="n">COCO_DETECTION_80_CLASSES_BBOX_ANCHORS</span> <span class="o">=</span> <span class="n">Anchors</span><span class="p">([[</span><span class="mi">10</span><span class="p">,</span> <span class="mi">13</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">30</span><span class="p">,</span> <span class="mi">33</span><span class="p">,</span> <span class="mi">23</span><span class="p">],</span>
    87. 

  87.                                                   <span class="p">[</span><span class="mi">30</span><span class="p">,</span> <span class="mi">61</span><span class="p">,</span> <span class="mi">62</span><span class="p">,</span> <span class="mi">45</span><span class="p">,</span> <span class="mi">59</span><span class="p">,</span> <span class="mi">119</span><span class="p">],</span>
    88. 

  88.                                                   <span class="p">[</span><span class="mi">116</span><span class="p">,</span> <span class="mi">90</span><span class="p">,</span> <span class="mi">156</span><span class="p">,</span> <span class="mi">198</span><span class="p">,</span> <span class="mi">373</span><span class="p">,</span> <span class="mi">326</span><span class="p">]],</span>
    89. 

  89.                                                   <span class="n">strides</span><span class="o">=</span><span class="p">[</span><span class="mi">8</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">32</span><span class="p">])</span>  <span class="c1"># output strides of all yolo outputs</span>
    90. 

  90. 

  91. <span class="n">DEFAULT_YOLOV5_ARCH_PARAMS</span> <span class="o">=</span> <span class="p">{</span>
    92. 

  92.     <span class="s1">&#39;anchors&#39;</span><span class="p">:</span> <span class="n">COCO_DETECTION_80_CLASSES_BBOX_ANCHORS</span><span class="p">,</span>  <span class="c1"># The sizes of the anchors predicted by the model</span>
    93. 

  93.     <span class="s1">&#39;num_classes&#39;</span><span class="p">:</span> <span class="mi">80</span><span class="p">,</span>  <span class="c1"># Number of classes to predict</span>
    94. 

  94.     <span class="s1">&#39;depth_mult_factor&#39;</span><span class="p">:</span> <span class="mf">1.0</span><span class="p">,</span>  <span class="c1"># depth multiplier for the entire model</span>
    95. 

  95.     <span class="s1">&#39;width_mult_factor&#39;</span><span class="p">:</span> <span class="mf">1.0</span><span class="p">,</span>  <span class="c1"># width multiplier for the entire model</span>
    96. 

  96.     <span class="s1">&#39;backbone_struct&#39;</span><span class="p">:</span> <span class="p">[</span><span class="mi">3</span><span class="p">,</span> <span class="mi">9</span><span class="p">,</span> <span class="mi">9</span><span class="p">,</span> <span class="mi">3</span><span class="p">],</span>  <span class="c1"># the number of blocks in every stage of the backbone</span>
    97. 

  97.     <span class="s1">&#39;channels_in&#39;</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span>  <span class="c1"># # of classes the model predicts</span>
    98. 

  98.     <span class="s1">&#39;skip_connections_dict&#39;</span><span class="p">:</span> <span class="p">{</span><span class="mi">12</span><span class="p">:</span> <span class="p">[</span><span class="mi">6</span><span class="p">],</span> <span class="mi">16</span><span class="p">:</span> <span class="p">[</span><span class="mi">4</span><span class="p">],</span> <span class="mi">19</span><span class="p">:</span> <span class="p">[</span><span class="mi">14</span><span class="p">],</span> <span class="mi">22</span><span class="p">:</span> <span class="p">[</span><span class="mi">10</span><span class="p">],</span> <span class="mi">24</span><span class="p">:</span> <span class="p">[</span><span class="mi">17</span><span class="p">,</span> <span class="mi">20</span><span class="p">]},</span>
    99. 

  99.     <span class="c1"># A dictionary defining skip connections. format is &#39;target: [source1, source2, ...]&#39;. Each item defines a skip</span>
    100. 

  100.     <span class="c1"># connection from all sources to the target according to the layer&#39;s index (count starts from the backbone)</span>
    101. 

  101.     <span class="s1">&#39;connection_layers_input_channel_size&#39;</span><span class="p">:</span> <span class="p">[</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">1024</span><span class="p">,</span> <span class="mi">512</span><span class="p">],</span>
    102. 

  102.     <span class="c1"># default number off channels for the connecting points between the backbone and the head</span>
    103. 

  103.     <span class="s1">&#39;fuse_conv_and_bn&#39;</span><span class="p">:</span> <span class="kc">False</span><span class="p">,</span>  <span class="c1"># Fuse sequential Conv + B.N layers into a single one</span>
    104. 

  104.     <span class="s1">&#39;add_nms&#39;</span><span class="p">:</span> <span class="kc">False</span><span class="p">,</span>  <span class="c1"># Add the NMS module to the computational graph</span>
    105. 

  105.     <span class="s1">&#39;nms_conf&#39;</span><span class="p">:</span> <span class="mf">0.25</span><span class="p">,</span>  <span class="c1"># When add_nms is True during NMS predictions with confidence lower than this will be discarded</span>
    106. 

  106.     <span class="s1">&#39;nms_iou&#39;</span><span class="p">:</span> <span class="mf">0.45</span><span class="p">,</span>  <span class="c1"># When add_nms is True IoU threshold for NMS algorithm</span>
    107. 

  107.                      <span class="c1"># (with smaller value more boxed will be considered &quot;the same&quot; and removed)</span>
    108. 

  108. <span class="p">}</span>
    109. 

  109. 

  110. 

  111. <div class="viewcode-block" id="YoloV5PostPredictionCallback"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoloV5PostPredictionCallback">[docs]</a><span class="k">class</span> <span class="nc">YoloV5PostPredictionCallback</span><span class="p">(</span><span class="n">DetectionPostPredictionCallback</span><span class="p">):</span>
    112. 

  112.     <span class="sd">&quot;&quot;&quot;Non-Maximum Suppression (NMS) module&quot;&quot;&quot;</span>
    113. 

  113. 

  114.     <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="n">conf</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.001</span><span class="p">,</span> <span class="n">iou</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.6</span><span class="p">,</span> <span class="n">classes</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="nb">int</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
    115. 

  115.                  <span class="n">nms_type</span><span class="p">:</span> <span class="n">NMS_Type</span> <span class="o">=</span> <span class="n">NMS_Type</span><span class="o">.</span><span class="n">ITERATIVE</span><span class="p">,</span> <span class="n">max_predictions</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">300</span><span class="p">):</span>
    116. 

  116.         <span class="sd">&quot;&quot;&quot;</span>
    117. 

  117. <span class="sd">        :param conf: confidence threshold</span>
    118. 

  118. <span class="sd">        :param iou: IoU threshold                                       (used in NMS_Type.ITERATIVE)</span>
    119. 

  119. <span class="sd">        :param classes: (optional list) filter by class                 (used in NMS_Type.ITERATIVE)</span>
    120. 

  120. <span class="sd">        :param nms_type: the type of nms to use (iterative or matrix)</span>
    121. 

  121. <span class="sd">        :param max_predictions: maximum number of boxes to output       (used in NMS_Type.MATRIX)</span>
    122. 

  122. <span class="sd">        &quot;&quot;&quot;</span>
    123. 

  123.         <span class="nb">super</span><span class="p">(</span><span class="n">YoloV5PostPredictionCallback</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>
    124. 

  124.         <span class="bp">self</span><span class="o">.</span><span class="n">conf</span> <span class="o">=</span> <span class="n">conf</span>
    125. 

  125.         <span class="bp">self</span><span class="o">.</span><span class="n">iou</span> <span class="o">=</span> <span class="n">iou</span>
    126. 

  126.         <span class="bp">self</span><span class="o">.</span><span class="n">classes</span> <span class="o">=</span> <span class="n">classes</span>
    127. 

  127.         <span class="bp">self</span><span class="o">.</span><span class="n">nms_type</span> <span class="o">=</span> <span class="n">nms_type</span>
    128. 

  128.         <span class="bp">self</span><span class="o">.</span><span class="n">max_predictions</span> <span class="o">=</span> <span class="n">max_predictions</span>
    129. 

  129. 

  130. <div class="viewcode-block" id="YoloV5PostPredictionCallback.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoloV5PostPredictionCallback.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">,</span> <span class="n">device</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="kc">None</span><span class="p">):</span>
    131. 

  131.         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">nms_type</span> <span class="o">==</span> <span class="n">NMS_Type</span><span class="o">.</span><span class="n">ITERATIVE</span><span class="p">:</span>
    132. 

  132.             <span class="k">return</span> <span class="n">non_max_suppression</span><span class="p">(</span><span class="n">x</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">conf_thres</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">conf</span><span class="p">,</span> <span class="n">iou_thres</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">iou</span><span class="p">,</span> <span class="n">classes</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">classes</span><span class="p">)</span>
    133. 

  133.         <span class="k">else</span><span class="p">:</span>
    134. 

  134.             <span class="k">return</span> <span class="n">matrix_non_max_suppression</span><span class="p">(</span><span class="n">x</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">conf_thres</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">conf</span><span class="p">,</span> <span class="n">max_num_of_detections</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">max_predictions</span><span class="p">)</span></div></div>
    135. 

  135. 

  136. 

  137. <div class="viewcode-block" id="Concat"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.Concat">[docs]</a><span class="k">class</span> <span class="nc">Concat</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
    138. 

  138.     <span class="sd">&quot;&quot;&quot; CONCATENATE A LIST OF TENSORS ALONG DIMENSION&quot;&quot;&quot;</span>
    139. 

  139. 

  140.     <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="n">dimension</span><span class="o">=</span><span class="mi">1</span><span class="p">):</span>
    141. 

  141.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    142. 

  142.         <span class="bp">self</span><span class="o">.</span><span class="n">dimension</span> <span class="o">=</span> <span class="n">dimension</span>
    143. 

  143. 

  144. <div class="viewcode-block" id="Concat.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.Concat.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    145. 

  145.         <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">dimension</span><span class="p">)</span></div></div>
    146. 

  146. 

  147. 

  148. <div class="viewcode-block" id="Detect"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.Detect">[docs]</a><span class="k">class</span> <span class="nc">Detect</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
    149. 

  149. 

  150.     <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="n">num_classes</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">anchors</span><span class="p">:</span> <span class="n">Anchors</span><span class="p">,</span> <span class="n">channels</span><span class="p">:</span> <span class="nb">list</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span>
    151. 

  151.                  <span class="n">width_mult_factor</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">):</span>
    152. 

  152.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    153. 

  153. 

  154.         <span class="c1"># CHANGING THE WIDTH OF EACH OF THE DETECTION LAYERS</span>
    155. 

  155.         <span class="n">channels</span> <span class="o">=</span> <span class="p">[</span><span class="n">width_multiplier</span><span class="p">(</span><span class="n">channel</span><span class="p">,</span> <span class="n">width_mult_factor</span><span class="p">)</span> <span class="k">for</span> <span class="n">channel</span> <span class="ow">in</span> <span class="n">channels</span><span class="p">]</span>
    156. 

  156. 

  157.         <span class="bp">self</span><span class="o">.</span><span class="n">num_classes</span> <span class="o">=</span> <span class="n">num_classes</span>
    158. 

  158.         <span class="bp">self</span><span class="o">.</span><span class="n">num_outputs</span> <span class="o">=</span> <span class="n">num_classes</span> <span class="o">+</span> <span class="mi">5</span>
    159. 

  159.         <span class="bp">self</span><span class="o">.</span><span class="n">detection_layers_num</span> <span class="o">=</span> <span class="n">anchors</span><span class="o">.</span><span class="n">detection_layers_num</span>
    160. 

  160.         <span class="bp">self</span><span class="o">.</span><span class="n">num_anchors</span> <span class="o">=</span> <span class="n">anchors</span><span class="o">.</span><span class="n">num_anchors</span>
    161. 

  161.         <span class="bp">self</span><span class="o">.</span><span class="n">grid</span> <span class="o">=</span> <span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">)]</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">detection_layers_num</span>  <span class="c1"># init grid</span>
    162. 

  162.         
    163. 

  163.         <span class="bp">self</span><span class="o">.</span><span class="n">register_buffer</span><span class="p">(</span><span class="s1">&#39;stride&#39;</span><span class="p">,</span> <span class="n">anchors</span><span class="o">.</span><span class="n">stride</span><span class="p">)</span>
    164. 

  164.         <span class="bp">self</span><span class="o">.</span><span class="n">register_buffer</span><span class="p">(</span><span class="s1">&#39;anchors&#39;</span><span class="p">,</span> <span class="n">anchors</span><span class="o">.</span><span class="n">anchors</span><span class="p">)</span>
    165. 

  165.         <span class="bp">self</span><span class="o">.</span><span class="n">register_buffer</span><span class="p">(</span><span class="s1">&#39;anchor_grid&#39;</span><span class="p">,</span> <span class="n">anchors</span><span class="o">.</span><span class="n">anchor_grid</span><span class="p">)</span>
    166. 

  166. 

  167.         <span class="bp">self</span><span class="o">.</span><span class="n">m</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ModuleList</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_outputs</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_anchors</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span> <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="n">channels</span><span class="p">)</span>  <span class="c1"># output conv</span>
    168. 

  168. 

  169. <div class="viewcode-block" id="Detect.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.Detect.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    170. 

  170.         <span class="n">z</span> <span class="o">=</span> <span class="p">[]</span>  <span class="c1"># inference output</span>
    171. 

  171.         <span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">detection_layers_num</span><span class="p">):</span>
    172. 

  172.             <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">m</span><span class="p">[</span><span class="n">i</span><span class="p">](</span><span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">])</span>  <span class="c1"># conv</span>
    173. 

  173.             <span class="n">bs</span><span class="p">,</span> <span class="n">_</span><span class="p">,</span> <span class="n">ny</span><span class="p">,</span> <span class="n">nx</span> <span class="o">=</span> <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">shape</span>  <span class="c1"># x(bs,255,20,20) to x(bs,3,20,20,85)</span>
    174. 

  174.             <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">bs</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_anchors</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_outputs</span><span class="p">,</span> <span class="n">ny</span><span class="p">,</span> <span class="n">nx</span><span class="p">)</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">contiguous</span><span class="p">()</span>
    175. 

  175. 

  176.             <span class="k">if</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span><span class="p">:</span>  <span class="c1"># inference</span>
    177. 

  177.                 <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">grid</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]</span> <span class="o">!=</span> <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]:</span>
    178. 

  178.                     <span class="bp">self</span><span class="o">.</span><span class="n">grid</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_make_grid</span><span class="p">(</span><span class="n">nx</span><span class="p">,</span> <span class="n">ny</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
    179. 

  179. 

  180.                 <span class="n">y</span> <span class="o">=</span> <span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">sigmoid</span><span class="p">()</span>
    181. 

  181.                 <span class="n">xy</span> <span class="o">=</span> <span class="p">(</span><span class="n">y</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">0</span><span class="p">:</span><span class="mi">2</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.</span> <span class="o">-</span> <span class="mf">0.5</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">grid</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">x</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">device</span><span class="p">))</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">stride</span><span class="p">[</span><span class="n">i</span><span class="p">]</span>  <span class="c1"># xy</span>
    182. 

  182.                 <span class="n">wh</span> <span class="o">=</span> <span class="p">(</span><span class="n">y</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">2</span><span class="p">:</span><span class="mi">4</span><span class="p">]</span> <span class="o">*</span> <span class="mi">2</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">anchor_grid</span><span class="p">[</span><span class="n">i</span><span class="p">]</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_anchors</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span>  <span class="c1"># wh</span>
    183. 

  183.                 <span class="n">y</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">([</span><span class="n">xy</span><span class="p">,</span> <span class="n">wh</span><span class="p">,</span> <span class="n">y</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">4</span><span class="p">:]],</span> <span class="n">dim</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span>
    184. 

  184.                 <span class="n">z</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">y</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">bs</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">num_outputs</span><span class="p">))</span>
    185. 

  185. 

  186.         <span class="k">return</span> <span class="n">x</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span> <span class="k">else</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="n">z</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="n">x</span><span class="p">)</span></div>
    187. 

  187. 

  188.     <span class="nd">@staticmethod</span>
    189. 

  189.     <span class="k">def</span> <span class="nf">_make_grid</span><span class="p">(</span><span class="n">nx</span><span class="o">=</span><span class="mi">20</span><span class="p">,</span> <span class="n">ny</span><span class="o">=</span><span class="mi">20</span><span class="p">):</span>
    190. 

  190.         <span class="n">yv</span><span class="p">,</span> <span class="n">xv</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">meshgrid</span><span class="p">([</span><span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">ny</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">nx</span><span class="p">)])</span>
    191. 

  191.         <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">((</span><span class="n">xv</span><span class="p">,</span> <span class="n">yv</span><span class="p">),</span> <span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">view</span><span class="p">((</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">ny</span><span class="p">,</span> <span class="n">nx</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span><span class="o">.</span><span class="n">float</span><span class="p">()</span></div>
    192. 

  192. 

  193. 

  194. <div class="viewcode-block" id="AbstractYoLoV5Backbone"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.AbstractYoLoV5Backbone">[docs]</a><span class="k">class</span> <span class="nc">AbstractYoLoV5Backbone</span><span class="p">:</span>
    195. 

  195.     <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="n">arch_params</span><span class="p">):</span>
    196. 

  196.         <span class="c1"># CREATE A LIST CONTAINING THE LAYERS TO EXTRACT FROM THE BACKBONE AND ADD THE FINAL LAYER</span>
    197. 

  197.         <span class="bp">self</span><span class="o">.</span><span class="n">_layer_idx_to_extract</span> <span class="o">=</span> <span class="p">[</span><span class="n">idx</span> <span class="k">for</span> <span class="n">sub_l</span> <span class="ow">in</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">skip_connections_dict</span><span class="o">.</span><span class="n">values</span><span class="p">()</span> <span class="k">for</span> <span class="n">idx</span> <span class="ow">in</span> <span class="n">sub_l</span><span class="p">]</span>
    198. 

  198.         <span class="bp">self</span><span class="o">.</span><span class="n">_layer_idx_to_extract</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span>
    199. 

  199. 

  200. <div class="viewcode-block" id="AbstractYoLoV5Backbone.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.AbstractYoLoV5Backbone.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    201. 

  201.         <span class="sd">&quot;&quot;&quot;:return A list, the length of self._modules_list containing the output of the layer if specified in</span>
    202. 

  202. <span class="sd">        self._layers_to_extract and None otherwise&quot;&quot;&quot;</span>
    203. 

  203.         <span class="n">extracted_intermediate_layers</span> <span class="o">=</span> <span class="p">[]</span>
    204. 

  204.         <span class="k">for</span> <span class="n">layer_idx</span><span class="p">,</span> <span class="n">layer_module</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">):</span>
    205. 

  205.             <span class="c1"># PREDICT THE NEXT LAYER&#39;S OUTPUT</span>
    206. 

  206.             <span class="n">x</span> <span class="o">=</span> <span class="n">layer_module</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
    207. 

  207.             <span class="c1"># IF INDICATED APPEND THE OUTPUT TO extracted_intermediate_layers O.W. APPEND None</span>
    208. 

  208.             <span class="n">extracted_intermediate_layers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">x</span><span class="p">)</span> <span class="k">if</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">_layer_idx_to_extract</span> \
    209. 

  209.                 <span class="k">else</span> <span class="n">extracted_intermediate_layers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
    210. 

  210. 

  211.         <span class="k">return</span> <span class="n">extracted_intermediate_layers</span></div></div>
    212. 

  212. 

  213. 

  214. <div class="viewcode-block" id="YoLoV5DarknetBackbone"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5DarknetBackbone">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5DarknetBackbone</span><span class="p">(</span><span class="n">AbstractYoLoV5Backbone</span><span class="p">,</span> <span class="n">CSPDarknet53</span><span class="p">):</span>
    215. 

  215.     <span class="sd">&quot;&quot;&quot;Implements the CSP_Darknet53 module and inherit the forward pass to extract layers indicated in arch_params&quot;&quot;&quot;</span>
    216. 

  216. 

  217.     <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="n">arch_params</span><span class="p">):</span>
    218. 

  218.         <span class="n">arch_params</span><span class="o">.</span><span class="n">backbone_mode</span> <span class="o">=</span> <span class="kc">True</span>
    219. 

  219.         <span class="n">CSPDarknet53</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">arch_params</span><span class="p">)</span>
    220. 

  220.         <span class="n">AbstractYoLoV5Backbone</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">arch_params</span><span class="p">)</span>
    221. 

  221. 

  222. <div class="viewcode-block" id="YoLoV5DarknetBackbone.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5DarknetBackbone.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    223. 

  223.         <span class="k">return</span> <span class="n">AbstractYoLoV5Backbone</span><span class="o">.</span><span class="n">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">)</span></div></div>
    224. 

  224. 

  225. 

  226. <div class="viewcode-block" id="YoLoV5Head"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Head">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5Head</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
    227. 

  227.     <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="n">arch_params</span><span class="p">):</span>
    228. 

  228.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    229. 

  229.         <span class="c1"># PARSE arch_params</span>
    230. 

  230.         <span class="n">num_classes</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">num_classes</span>
    231. 

  231.         <span class="n">depth_mult_factor</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">depth_mult_factor</span>
    232. 

  232.         <span class="n">width_mult_factor</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span>
    233. 

  233.         <span class="n">anchors</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">anchors</span>
    234. 

  234.         <span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">skip_connections_dict</span>
    235. 

  235.         <span class="c1"># FLATTEN THE SOURCE LIST INTO A LIST OF INDICES</span>
    236. 

  236.         <span class="bp">self</span><span class="o">.</span><span class="n">_layer_idx_to_extract</span> <span class="o">=</span> <span class="p">[</span><span class="n">idx</span> <span class="k">for</span> <span class="n">sub_l</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span><span class="o">.</span><span class="n">values</span><span class="p">()</span> <span class="k">for</span> <span class="n">idx</span> <span class="ow">in</span> <span class="n">sub_l</span><span class="p">]</span>
    237. 

  237. 

  238.         <span class="c1"># GET THREE CONNECTING POINTS CHANNEL INPUT SIZE</span>
    239. 

  239.         <span class="n">connector</span> <span class="o">=</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">connection_layers_input_channel_size</span>
    240. 

  240. 

  241.         <span class="n">width_mult</span> <span class="o">=</span> <span class="k">lambda</span> <span class="n">channels</span><span class="p">:</span> <span class="n">width_multiplier</span><span class="p">(</span><span class="n">channels</span><span class="p">,</span> <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span><span class="p">)</span>
    242. 

  242. 

  243.         <span class="c1"># THE MODULES LIST IS APPROACHABLE FROM &quot;OUTSIDE THE CLASS - SO WE CAN CHANGE IT&#39;S STRUCTURE&quot;</span>
    244. 

  244.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">ModuleList</span><span class="p">()</span>
    245. 

  245. 

  246.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Conv</span><span class="p">(</span><span class="n">width_mult</span><span class="p">(</span><span class="n">connector</span><span class="p">[</span><span class="mi">0</span><span class="p">]),</span> <span class="n">width_mult</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">))</span>  <span class="c1"># 10</span>
    247. 

  247.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Upsample</span><span class="p">(</span><span class="kc">None</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="s1">&#39;nearest&#39;</span><span class="p">))</span>  <span class="c1"># 11</span>
    248. 

  248.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Concat</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>  <span class="c1"># 12</span>
    249. 

  249.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">BottleneckCSP</span><span class="p">(</span><span class="n">connector</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="n">width_mult_factor</span><span class="o">=</span><span class="n">width_mult_factor</span><span class="p">,</span>
    250. 

  250.                                                 <span class="n">depth_mult_factor</span><span class="o">=</span><span class="n">depth_mult_factor</span><span class="p">))</span>  <span class="c1"># 13</span>
    251. 

  251.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Conv</span><span class="p">(</span><span class="n">width_mult</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span> <span class="n">width_mult</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span> <span class="mi">1</span><span class="p">))</span>  <span class="c1"># 14</span>
    252. 

  252.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Upsample</span><span class="p">(</span><span class="kc">None</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="s1">&#39;nearest&#39;</span><span class="p">))</span>  <span class="c1"># 15</span>
    253. 

  253.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Concat</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>  <span class="c1"># 16</span>
    254. 

  254.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">BottleneckCSP</span><span class="p">(</span><span class="n">connector</span><span class="p">[</span><span class="mi">2</span><span class="p">],</span> <span class="mi">256</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="n">width_mult_factor</span><span class="o">=</span><span class="n">width_mult_factor</span><span class="p">,</span>
    255. 

  255.                                                 <span class="n">depth_mult_factor</span><span class="o">=</span><span class="n">depth_mult_factor</span><span class="p">))</span>  <span class="c1"># 17</span>
    256. 

  256.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Conv</span><span class="p">(</span><span class="n">width_mult</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span> <span class="n">width_mult</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span>  <span class="c1"># 18</span>
    257. 

  257.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Concat</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>  <span class="c1"># 19</span>
    258. 

  258.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">BottleneckCSP</span><span class="p">(</span><span class="mi">512</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="n">width_mult_factor</span><span class="o">=</span><span class="n">width_mult_factor</span><span class="p">,</span>
    259. 

  259.                                                 <span class="n">depth_mult_factor</span><span class="o">=</span><span class="n">depth_mult_factor</span><span class="p">))</span>  <span class="c1"># 20</span>
    260. 

  260.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Conv</span><span class="p">(</span><span class="n">width_mult</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span> <span class="n">width_mult</span><span class="p">(</span><span class="mi">512</span><span class="p">),</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span>  <span class="c1"># 21</span>
    261. 

  261.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Concat</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span>  <span class="c1"># 22</span>
    262. 

  262.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">BottleneckCSP</span><span class="p">(</span><span class="mi">1024</span><span class="p">,</span> <span class="mi">1024</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="kc">False</span><span class="p">,</span> <span class="n">width_mult_factor</span><span class="o">=</span><span class="n">width_mult_factor</span><span class="p">,</span>
    263. 

  263.                                                 <span class="n">depth_mult_factor</span><span class="o">=</span><span class="n">depth_mult_factor</span><span class="p">))</span>  <span class="c1"># 23</span>
    264. 

  264. 

  265.         <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Detect</span><span class="p">(</span><span class="n">num_classes</span><span class="p">,</span> <span class="n">anchors</span><span class="p">,</span> <span class="n">channels</span><span class="o">=</span><span class="p">[</span><span class="mi">256</span><span class="p">,</span> <span class="mi">512</span><span class="p">,</span> <span class="mi">1024</span><span class="p">],</span>
    266. 

  266.                                          <span class="n">width_mult_factor</span><span class="o">=</span><span class="n">width_mult_factor</span><span class="p">))</span>  <span class="c1"># 24</span>
    267. 

  267. 

  268. <div class="viewcode-block" id="YoLoV5Head.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Head.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">intermediate_output</span><span class="p">):</span>
    269. 

  269.         <span class="sd">&quot;&quot;&quot;</span>
    270. 

  270. <span class="sd">        :param intermediate_output: A list of the intermediate prediction of layers specified in the</span>
    271. 

  271. <span class="sd">        self._inter_layer_idx_to_extract from the Backbone</span>
    272. 

  272. <span class="sd">        &quot;&quot;&quot;</span>
    273. 

  273.         <span class="c1"># COUNT THE NUMBER OF LAYERS IN THE BACKBONE TO CONTINUE THE COUNTER</span>
    274. 

  274.         <span class="n">num_layers_in_backbone</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">intermediate_output</span><span class="p">)</span>
    275. 

  275.         <span class="c1"># INPUT TO HEAD IS THE LAST ELEMENT OF THE BACKBONE&#39;S OUTPUT</span>
    276. 

  276.         <span class="n">out</span> <span class="o">=</span> <span class="n">intermediate_output</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>
    277. 

  277.         <span class="c1"># RUN OVER THE MODULE LIST WITHOUT THE FINAL LAYER &amp; START COUNTER FROM THE END OF THE BACKBONE</span>
    278. 

  278.         <span class="k">for</span> <span class="n">layer_idx</span><span class="p">,</span> <span class="n">layer_module</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">[:</span><span class="o">-</span><span class="mi">1</span><span class="p">],</span> <span class="n">start</span><span class="o">=</span><span class="n">num_layers_in_backbone</span><span class="p">):</span>
    279. 

  279.             <span class="c1"># IF THE LAYER APPEARS IN THE KEYS IT INSERT THE PRECIOUS OUTPUT AND THE INDICATED SKIP CONNECTIONS</span>
    280. 

  280.             <span class="n">out</span> <span class="o">=</span> <span class="n">layer_module</span><span class="p">([</span><span class="n">out</span><span class="p">,</span> <span class="n">intermediate_output</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span><span class="p">[</span><span class="n">layer_idx</span><span class="p">][</span><span class="mi">0</span><span class="p">]]])</span> \
    281. 

  281.                 <span class="k">if</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span><span class="o">.</span><span class="n">keys</span><span class="p">()</span> <span class="k">else</span> <span class="n">layer_module</span><span class="p">(</span><span class="n">out</span><span class="p">)</span>
    282. 

  282. 

  283.             <span class="c1"># IF INDICATED APPEND THE OUTPUT TO inter_layer_idx_to_extract O.W. APPEND None</span>
    284. 

  284.             <span class="n">intermediate_output</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">out</span><span class="p">)</span> <span class="k">if</span> <span class="n">layer_idx</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">_layer_idx_to_extract</span> \
    285. 

  285.                 <span class="k">else</span> <span class="n">intermediate_output</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
    286. 

  286. 

  287.         <span class="c1"># INSERT THE REMAINING LAYERS INTO THE Detect LAYER</span>
    288. 

  288.         <span class="n">last_idx</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">_modules_list</span><span class="p">)</span> <span class="o">+</span> <span class="n">num_layers_in_backbone</span> <span class="o">-</span> <span class="mi">1</span>
    289. 

  289.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]([</span><span class="n">intermediate_output</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span><span class="p">[</span><span class="n">last_idx</span><span class="p">][</span><span class="mi">0</span><span class="p">]],</span>
    290. 

  290.                                        <span class="n">intermediate_output</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">_skip_connections_dict</span><span class="p">[</span><span class="n">last_idx</span><span class="p">][</span><span class="mi">1</span><span class="p">]],</span>
    291. 

  291.                                        <span class="n">out</span><span class="p">])</span></div></div>
    292. 

  292. 

  293. 

  294. <div class="viewcode-block" id="YoLoV5Base"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5Base</span><span class="p">(</span><span class="n">SgModule</span><span class="p">):</span>
    295. 

  295.     <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="n">backbone</span><span class="p">:</span> <span class="n">Type</span><span class="p">[</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">],</span> <span class="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">,</span> <span class="n">initialize_module</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">):</span>
    296. 

  296.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
    297. 

  297.         <span class="c1"># DEFAULT PARAMETERS TO BE OVERWRITTEN BY DUPLICATES THAT APPEAR IN arch_params</span>
    298. 

  298.         <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span> <span class="o">=</span> <span class="n">HpmStruct</span><span class="p">(</span><span class="o">**</span><span class="n">DEFAULT_YOLOV5_ARCH_PARAMS</span><span class="p">)</span>
    299. 

  299.         <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">override</span><span class="p">(</span><span class="o">**</span><span class="n">arch_params</span><span class="o">.</span><span class="n">to_dict</span><span class="p">())</span>
    300. 

  300. 

  301.         <span class="bp">self</span><span class="o">.</span><span class="n">num_classes</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">num_classes</span>
    302. 

  302.         <span class="c1"># THE MODEL&#39;S MODULES</span>
    303. 

  303.         <span class="bp">self</span><span class="o">.</span><span class="n">_backbone</span> <span class="o">=</span> <span class="n">backbone</span><span class="p">(</span><span class="n">arch_params</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="p">)</span>
    304. 

  304.         <span class="bp">self</span><span class="o">.</span><span class="n">_nms</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Identity</span><span class="p">()</span>
    305. 

  305. 

  306.         <span class="c1"># A FLAG TO DEFINE augment_forward IN INFERENCE</span>
    307. 

  307.         <span class="bp">self</span><span class="o">.</span><span class="n">augmented_inference</span> <span class="o">=</span> <span class="kc">False</span>
    308. 

  308. 

  309.         <span class="c1"># RUN SPECIFIC INITIALIZATION OF YOLO-V5</span>
    310. 

  310.         <span class="k">if</span> <span class="n">initialize_module</span><span class="p">:</span>
    311. 

  311.             <span class="bp">self</span><span class="o">.</span><span class="n">_head</span> <span class="o">=</span> <span class="n">YoLoV5Head</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="p">)</span>
    312. 

  312.             <span class="bp">self</span><span class="o">.</span><span class="n">_initialize_module</span><span class="p">()</span>
    313. 

  313. 

  314. <div class="viewcode-block" id="YoLoV5Base.forward"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.forward">[docs]</a>    <span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    315. 

  315.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_augment_forward</span><span class="p">(</span><span class="n">x</span><span class="p">)</span> <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">augmented_inference</span> <span class="k">else</span> <span class="bp">self</span><span class="o">.</span><span class="n">_forward_once</span><span class="p">(</span><span class="n">x</span><span class="p">)</span></div>
    316. 

  316. 

  317.     <span class="k">def</span> <span class="nf">_forward_once</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    318. 

  318.         <span class="n">out</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_backbone</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
    319. 

  319.         <span class="n">out</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_head</span><span class="p">(</span><span class="n">out</span><span class="p">)</span>
    320. 

  320.         <span class="c1"># THIS HAS NO EFFECT IF add_nms() WAS NOT DONE</span>
    321. 

  321.         <span class="n">out</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_nms</span><span class="p">(</span><span class="n">out</span><span class="p">)</span>
    322. 

  322.         <span class="k">return</span> <span class="n">out</span>
    323. 

  323. 

  324.     <span class="k">def</span> <span class="nf">_augment_forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
    325. 

  325.         <span class="sd">&quot;&quot;&quot;Multi-scale forward pass&quot;&quot;&quot;</span>
    326. 

  326.         <span class="n">img_size</span> <span class="o">=</span> <span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">:]</span>  <span class="c1"># height, width</span>
    327. 

  327.         <span class="n">s</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mf">0.83</span><span class="p">,</span> <span class="mf">0.67</span><span class="p">]</span>  <span class="c1"># scales</span>
    328. 

  328.         <span class="n">f</span> <span class="o">=</span> <span class="p">[</span><span class="kc">None</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="kc">None</span><span class="p">]</span>  <span class="c1"># flips (2-ud, 3-lr)</span>
    329. 

  329.         <span class="n">y</span> <span class="o">=</span> <span class="p">[]</span>  <span class="c1"># outputs</span>
    330. 

  330.         <span class="k">for</span> <span class="n">si</span><span class="p">,</span> <span class="n">fi</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="n">f</span><span class="p">):</span>
    331. 

  331.             <span class="n">xi</span> <span class="o">=</span> <span class="n">scale_img</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">flip</span><span class="p">(</span><span class="n">fi</span><span class="p">)</span> <span class="k">if</span> <span class="n">fi</span> <span class="k">else</span> <span class="n">x</span><span class="p">,</span> <span class="n">si</span><span class="p">)</span>
    332. 

  332.             <span class="n">yi</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_forward_once</span><span class="p">(</span><span class="n">xi</span><span class="p">)[</span><span class="mi">0</span><span class="p">]</span>  <span class="c1"># forward</span>
    333. 

  333.             <span class="n">yi</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="p">:</span><span class="mi">4</span><span class="p">]</span> <span class="o">/=</span> <span class="n">si</span>  <span class="c1"># de-scale</span>
    334. 

  334.             <span class="k">if</span> <span class="n">fi</span> <span class="o">==</span> <span class="mi">2</span><span class="p">:</span>
    335. 

  335.                 <span class="n">yi</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span> <span class="o">=</span> <span class="n">img_size</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">-</span> <span class="n">yi</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>  <span class="c1"># de-flip ud</span>
    336. 

  336.             <span class="k">elif</span> <span class="n">fi</span> <span class="o">==</span> <span class="mi">3</span><span class="p">:</span>
    337. 

  337.                 <span class="n">yi</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="n">img_size</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="o">-</span> <span class="n">yi</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="mi">0</span><span class="p">]</span>  <span class="c1"># de-flip lr</span>
    338. 

  338.             <span class="n">y</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">yi</span><span class="p">)</span>
    339. 

  339.         <span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="n">y</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="kc">None</span>  <span class="c1"># augmented inference, train</span>
    340. 

  340. 

  341. <div class="viewcode-block" id="YoLoV5Base.load_state_dict"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.load_state_dict">[docs]</a>    <span class="k">def</span> <span class="nf">load_state_dict</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">state_dict</span><span class="p">,</span> <span class="n">strict</span><span class="o">=</span><span class="kc">True</span><span class="p">):</span>
    342. 

  342.         <span class="k">try</span><span class="p">:</span>
    343. 

  343.             <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">state_dict</span><span class="p">,</span> <span class="n">strict</span><span class="p">)</span>
    344. 

  344.         <span class="k">except</span> <span class="ne">RuntimeError</span> <span class="k">as</span> <span class="n">e</span><span class="p">:</span>
    345. 

  345.             <span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="sa">f</span><span class="s2">&quot;Got exception </span><span class="si">{</span><span class="n">e</span><span class="si">}</span><span class="s2">, if a mismatch between expected and given state_dict keys exist, &quot;</span>
    346. 

  346.                                <span class="sa">f</span><span class="s2">&quot;checkpoint may have been saved after fusing conv and bn. use fuse_conv_bn before loading.&quot;</span><span class="p">)</span></div>
    347. 

  347. 

  348.     <span class="k">def</span> <span class="nf">_initialize_module</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    349. 

  349.         <span class="bp">self</span><span class="o">.</span><span class="n">_check_strides_and_anchors</span><span class="p">()</span>
    350. 

  350.         <span class="bp">self</span><span class="o">.</span><span class="n">_initialize_biases</span><span class="p">()</span>
    351. 

  351.         <span class="bp">self</span><span class="o">.</span><span class="n">_initialize_weights</span><span class="p">()</span>
    352. 

  352.         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">add_nms</span><span class="p">:</span>
    353. 

  353.             <span class="n">nms_conf</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">nms_conf</span>
    354. 

  354.             <span class="n">nms_iou</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">nms_iou</span>
    355. 

  355.             <span class="bp">self</span><span class="o">.</span><span class="n">_nms</span> <span class="o">=</span> <span class="n">YoloV5PostPredictionCallback</span><span class="p">(</span><span class="n">nms_conf</span><span class="p">,</span> <span class="n">nms_iou</span><span class="p">)</span>
    356. 

  356. 

  357. <div class="viewcode-block" id="YoLoV5Base.update_param_groups"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.update_param_groups">[docs]</a>    <span class="k">def</span> <span class="nf">update_param_groups</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">param_groups</span><span class="p">:</span> <span class="nb">list</span><span class="p">,</span> <span class="n">lr</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">epoch</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="nb">iter</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    358. 

  358.                             <span class="n">training_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">,</span> <span class="n">total_batch</span><span class="p">:</span> <span class="nb">int</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">list</span><span class="p">:</span>
    359. 

  359. 

  360.         <span class="n">lr_warmup_epochs</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">training_params</span><span class="p">,</span> <span class="s1">&#39;lr_warmup_epochs&#39;</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span>
    361. 

  361.         <span class="k">if</span> <span class="n">epoch</span> <span class="o">&lt;</span> <span class="n">lr_warmup_epochs</span> <span class="ow">and</span> <span class="nb">iter</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
    362. 

  362.             <span class="c1"># OVERRIDE THE lr FROM SgModel WITH initial_lr, SINCE SgModel MANIPULATE THE ORIGINAL VALUE</span>
    363. 

  363.             <span class="n">print_once</span><span class="p">(</span><span class="s1">&#39;Using Yolo v5 warm-up lr (overriding ModelBase lr function)&#39;</span><span class="p">)</span>
    364. 

  364.             <span class="n">lr</span> <span class="o">=</span> <span class="n">training_params</span><span class="o">.</span><span class="n">initial_lr</span>
    365. 

  365.             <span class="n">momentum</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">training_params</span><span class="o">.</span><span class="n">optimizer_params</span><span class="p">,</span> <span class="s1">&#39;momentum&#39;</span><span class="p">)</span>
    366. 

  366.             <span class="n">warmup_momentum</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">training_params</span><span class="p">,</span> <span class="s1">&#39;warmup_momentum&#39;</span><span class="p">,</span> <span class="n">momentum</span><span class="p">)</span>
    367. 

  367.             <span class="n">warmup_bias_lr</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">training_params</span><span class="p">,</span> <span class="s1">&#39;warmup_bias_lr&#39;</span><span class="p">,</span> <span class="n">lr</span><span class="p">)</span>
    368. 

  368.             <span class="n">nw</span> <span class="o">=</span> <span class="n">lr_warmup_epochs</span> <span class="o">*</span> <span class="n">total_batch</span>
    369. 

  369.             <span class="n">ni</span> <span class="o">=</span> <span class="n">epoch</span> <span class="o">*</span> <span class="n">total_batch</span> <span class="o">+</span> <span class="nb">iter</span>
    370. 

  370.             <span class="n">xi</span> <span class="o">=</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="n">nw</span><span class="p">]</span>  <span class="c1"># x interp</span>
    371. 

  371.             <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="n">param_groups</span><span class="p">:</span>
    372. 

  372.                 <span class="c1"># BIAS LR FALLS FROM 0.1 TO LR0, ALL OTHER LRS RISE FROM 0.0 TO LR0</span>
    373. 

  373.                 <span class="n">x</span><span class="p">[</span><span class="s1">&#39;lr&#39;</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">interp</span><span class="p">(</span><span class="n">ni</span><span class="p">,</span> <span class="n">xi</span><span class="p">,</span> <span class="p">[</span><span class="n">warmup_bias_lr</span> <span class="k">if</span> <span class="n">x</span><span class="p">[</span><span class="s1">&#39;name&#39;</span><span class="p">]</span> <span class="o">==</span> <span class="s1">&#39;bias&#39;</span> <span class="k">else</span> <span class="mf">0.0</span><span class="p">,</span> <span class="n">lr</span><span class="p">])</span>
    374. 

  374.                 <span class="k">if</span> <span class="s1">&#39;momentum&#39;</span> <span class="ow">in</span> <span class="n">x</span><span class="p">:</span>
    375. 

  375.                     <span class="n">x</span><span class="p">[</span><span class="s1">&#39;momentum&#39;</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">interp</span><span class="p">(</span><span class="n">ni</span><span class="p">,</span> <span class="n">xi</span><span class="p">,</span> <span class="p">[</span><span class="n">warmup_momentum</span><span class="p">,</span> <span class="n">momentum</span><span class="p">])</span>
    376. 

  376.             <span class="k">return</span> <span class="n">param_groups</span>
    377. 

  377.         <span class="k">else</span><span class="p">:</span>
    378. 

  378.             <span class="k">return</span> <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="n">update_param_groups</span><span class="p">(</span><span class="n">param_groups</span><span class="p">,</span> <span class="n">lr</span><span class="p">,</span> <span class="n">epoch</span><span class="p">,</span> <span class="nb">iter</span><span class="p">,</span> <span class="n">training_params</span><span class="p">,</span> <span class="n">total_batch</span><span class="p">)</span></div>
    379. 

  379. 

  380.     <span class="k">def</span> <span class="nf">_check_strides_and_anchors</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    381. 

  381.         <span class="n">m</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_head</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>  <span class="c1"># Detect()</span>
    382. 

  382.         <span class="c1"># Do inference in train mode on a dummy image to get output stride of each head output layer</span>
    383. 

  383.         <span class="n">s</span> <span class="o">=</span> <span class="mi">128</span>  <span class="c1"># twice the minimum acceptable image size </span>
    384. 

  384.         <span class="n">dummy_input</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span><span class="o">.</span><span class="n">channels_in</span><span class="p">,</span> <span class="n">s</span><span class="p">,</span> <span class="n">s</span><span class="p">)</span>
    385. 

  385.         <span class="n">stride</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">([</span><span class="n">s</span> <span class="o">/</span> <span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">]</span> <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">_forward_once</span><span class="p">(</span><span class="n">dummy_input</span><span class="p">)])</span>
    386. 

  386.         <span class="k">if</span> <span class="ow">not</span> <span class="n">torch</span><span class="o">.</span><span class="n">equal</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">stride</span><span class="p">,</span> <span class="n">stride</span><span class="p">):</span>
    387. 

  387.             <span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s1">&#39;Provided anchor strides do not match the model strides&#39;</span><span class="p">)</span>
    388. 

  388.         <span class="n">check_anchor_order</span><span class="p">(</span><span class="n">m</span><span class="p">)</span>
    389. 

  389. 

  390.         <span class="bp">self</span><span class="o">.</span><span class="n">register_buffer</span><span class="p">(</span><span class="s1">&#39;stride&#39;</span><span class="p">,</span> <span class="n">m</span><span class="o">.</span><span class="n">stride</span><span class="p">)</span>  <span class="c1"># USED ONLY FOR CONVERSION</span>
    391. 

  391. 

  392.     <span class="k">def</span> <span class="nf">_initialize_biases</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">cf</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
    393. 

  393.         <span class="sd">&quot;&quot;&quot;initialize biases into Detect(), cf is class frequency&quot;&quot;&quot;</span>
    394. 

  394.         <span class="c1"># TODO: UNDERSTAND WHAT IS THIS cf AND IF WE NEED IT</span>
    395. 

  395.         <span class="c1"># cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.</span>
    396. 

  396.         <span class="n">m</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">_head</span><span class="o">.</span><span class="n">_modules_list</span><span class="p">[</span><span class="o">-</span><span class="mi">1</span><span class="p">]</span>  <span class="c1"># Detect() module</span>
    397. 

  397.         <span class="k">for</span> <span class="n">mi</span><span class="p">,</span> <span class="n">s</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">m</span><span class="p">,</span> <span class="n">m</span><span class="o">.</span><span class="n">stride</span><span class="p">):</span>  <span class="c1"># from</span>
    398. 

  398.             <span class="n">b</span> <span class="o">=</span> <span class="n">mi</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">num_anchors</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span>  <span class="c1"># conv.bias(255) to (3,85)</span>
    399. 

  399.             <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span>
    400. 

  400.                 <span class="n">b</span><span class="p">[:,</span> <span class="mi">4</span><span class="p">]</span> <span class="o">+=</span> <span class="n">math</span><span class="o">.</span><span class="n">log</span><span class="p">(</span><span class="mi">8</span> <span class="o">/</span> <span class="p">(</span><span class="mi">640</span> <span class="o">/</span> <span class="n">s</span><span class="p">)</span> <span class="o">**</span> <span class="mi">2</span><span class="p">)</span>  <span class="c1"># obj (8 objects per 640 image)</span>
    401. 

  401.                 <span class="n">b</span><span class="p">[:,</span> <span class="mi">5</span><span class="p">:]</span> <span class="o">+=</span> <span class="n">math</span><span class="o">.</span><span class="n">log</span><span class="p">(</span><span class="mf">0.6</span> <span class="o">/</span> <span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">num_classes</span> <span class="o">-</span> <span class="mf">0.99</span><span class="p">))</span> <span class="k">if</span> <span class="n">cf</span> <span class="ow">is</span> <span class="kc">None</span> <span class="k">else</span> <span class="n">torch</span><span class="o">.</span><span class="n">log</span><span class="p">(</span><span class="n">cf</span> <span class="o">/</span> <span class="n">cf</span><span class="o">.</span><span class="n">sum</span><span class="p">())</span>  <span class="c1"># cls</span>
    402. 

  402.             <span class="n">mi</span><span class="o">.</span><span class="n">bias</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">b</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">),</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    403. 

  403. 

  404.     <span class="k">def</span> <span class="nf">_initialize_weights</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    405. 

  405.         <span class="k">for</span> <span class="n">m</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">modules</span><span class="p">():</span>
    406. 

  406.             <span class="n">t</span> <span class="o">=</span> <span class="nb">type</span><span class="p">(</span><span class="n">m</span><span class="p">)</span>
    407. 

  407.             <span class="k">if</span> <span class="n">t</span> <span class="ow">is</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">:</span>
    408. 

  408.                 <span class="k">pass</span>  <span class="c1"># nn.init.kaiming_normal_(m.weight, mode=&#39;fan_out&#39;, nonlinearity=&#39;relu&#39;)</span>
    409. 

  409.             <span class="k">elif</span> <span class="n">t</span> <span class="ow">is</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">:</span>
    410. 

  410.                 <span class="n">m</span><span class="o">.</span><span class="n">eps</span> <span class="o">=</span> <span class="mf">1e-3</span>
    411. 

  411.                 <span class="n">m</span><span class="o">.</span><span class="n">momentum</span> <span class="o">=</span> <span class="mf">0.03</span>
    412. 

  412.             <span class="k">elif</span> <span class="n">t</span> <span class="ow">in</span> <span class="p">[</span><span class="n">nn</span><span class="o">.</span><span class="n">LeakyReLU</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">ReLU6</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Hardswish</span><span class="p">]:</span>
    413. 

  413.                 <span class="n">m</span><span class="o">.</span><span class="n">inplace</span> <span class="o">=</span> <span class="kc">True</span>
    414. 

  414. 

  415. <div class="viewcode-block" id="YoLoV5Base.initialize_param_groups"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.initialize_param_groups">[docs]</a>    <span class="k">def</span> <span class="nf">initialize_param_groups</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">lr</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="n">training_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">list</span><span class="p">:</span>
    416. 

  416.         <span class="sd">&quot;&quot;&quot;</span>
    417. 

  417. <span class="sd">        initialize_optimizer_for_model_param_groups - Initializes the weights of the optimizer</span>
    418. 

  418. <span class="sd">                                                      adds weight decay  *Only* to the Conv2D layers</span>
    419. 

  419. <span class="sd">            :param optimizer_cls:   The nn.optim (optimizer class) to initialize</span>
    420. 

  420. <span class="sd">            :param lr:              lr to set for the optimizer</span>
    421. 

  421. <span class="sd">            :param training_params:</span>
    422. 

  422. <span class="sd">            :return: The optimizer, initialized with the relevant param groups</span>
    423. 

  423. <span class="sd">        &quot;&quot;&quot;</span>
    424. 

  424.         <span class="n">optimizer_params</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">training_params</span><span class="p">,</span> <span class="s1">&#39;optimizer_params&#39;</span><span class="p">)</span>
    425. 

  425.         <span class="c1"># OPTIMIZER PARAMETER GROUPS</span>
    426. 

  426.         <span class="n">default_param_group</span><span class="p">,</span> <span class="n">weight_decay_param_group</span><span class="p">,</span> <span class="n">biases_param_group</span> <span class="o">=</span> <span class="p">[],</span> <span class="p">[],</span> <span class="p">[]</span>
    427. 

  427. 

  428.         <span class="k">for</span> <span class="n">name</span><span class="p">,</span> <span class="n">m</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">named_modules</span><span class="p">():</span>
    429. 

  429.             <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="s1">&#39;bias&#39;</span><span class="p">)</span> <span class="ow">and</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">):</span>  <span class="c1"># bias</span>
    430. 

  430.                 <span class="n">biases_param_group</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">name</span><span class="p">,</span> <span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="p">))</span>
    431. 

  431.             <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">):</span>  <span class="c1"># weight (no decay)</span>
    432. 

  432.                 <span class="n">default_param_group</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">name</span><span class="p">,</span> <span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">))</span>
    433. 

  433.             <span class="k">elif</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="s1">&#39;weight&#39;</span><span class="p">)</span> <span class="ow">and</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">):</span>  <span class="c1"># weight (with decay)</span>
    434. 

  434.                 <span class="n">weight_decay_param_group</span><span class="o">.</span><span class="n">append</span><span class="p">((</span><span class="n">name</span><span class="p">,</span> <span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">))</span>
    435. 

  435. 

  436.         <span class="c1"># EXTRACT weight_decay FROM THE optimizer_params IN ORDER TO ASSIGN THEM MANUALLY</span>
    437. 

  437.         <span class="n">weight_decay</span> <span class="o">=</span> <span class="n">optimizer_params</span><span class="o">.</span><span class="n">pop</span><span class="p">(</span><span class="s1">&#39;weight_decay&#39;</span><span class="p">)</span> <span class="k">if</span> <span class="s1">&#39;weight_decay&#39;</span> <span class="ow">in</span> <span class="n">optimizer_params</span><span class="o">.</span><span class="n">keys</span><span class="p">()</span> <span class="k">else</span> <span class="mi">0</span>
    438. 

  438.         <span class="n">param_groups</span> <span class="o">=</span> <span class="p">[{</span><span class="s1">&#39;named_params&#39;</span><span class="p">:</span> <span class="n">default_param_group</span><span class="p">,</span> <span class="s1">&#39;lr&#39;</span><span class="p">:</span> <span class="n">lr</span><span class="p">,</span> <span class="o">**</span><span class="n">optimizer_params</span><span class="p">,</span> <span class="s1">&#39;name&#39;</span><span class="p">:</span> <span class="s1">&#39;default&#39;</span><span class="p">},</span>
    439. 

  439.                         <span class="p">{</span><span class="s1">&#39;named_params&#39;</span><span class="p">:</span> <span class="n">weight_decay_param_group</span><span class="p">,</span> <span class="s1">&#39;weight_decay&#39;</span><span class="p">:</span> <span class="n">weight_decay</span><span class="p">,</span> <span class="s1">&#39;name&#39;</span><span class="p">:</span> <span class="s1">&#39;wd&#39;</span><span class="p">},</span>
    440. 

  440.                         <span class="p">{</span><span class="s1">&#39;named_params&#39;</span><span class="p">:</span> <span class="n">biases_param_group</span><span class="p">,</span> <span class="s1">&#39;name&#39;</span><span class="p">:</span> <span class="s1">&#39;bias&#39;</span><span class="p">}]</span>
    441. 

  441. 

  442.         <span class="c1"># Assert that all parameters were added to optimizer param groups</span>
    443. 

  443.         <span class="n">params_total</span> <span class="o">=</span> <span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">parameters</span><span class="p">())</span>
    444. 

  444.         <span class="n">optimizer_params_total</span> <span class="o">=</span> <span class="nb">sum</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="n">param_groups</span> <span class="k">for</span> <span class="n">_</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">g</span><span class="p">[</span><span class="s1">&#39;named_params&#39;</span><span class="p">])</span>
    445. 

  445.         <span class="k">assert</span> <span class="n">params_total</span> <span class="o">==</span> <span class="n">optimizer_params_total</span><span class="p">,</span> \
    446. 

  446.             <span class="sa">f</span><span class="s2">&quot;Parameters </span><span class="si">{</span><span class="p">[</span><span class="n">n</span> <span class="k">for</span> <span class="n">n</span><span class="p">,</span> <span class="n">_</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()</span> <span class="k">if</span> <span class="s1">&#39;weight&#39;</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">n</span> <span class="ow">and</span> <span class="s1">&#39;bias&#39;</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">n</span><span class="p">]</span><span class="si">}</span><span class="s2"> &quot;</span> \
    447. 

  447.             <span class="sa">f</span><span class="s2">&quot;weren&#39;t added to optimizer param groups&quot;</span>
    448. 

  448. 

  449.         <span class="k">return</span> <span class="n">param_groups</span></div>
    450. 

  450. 

  451. <div class="viewcode-block" id="YoLoV5Base.prep_model_for_conversion"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.prep_model_for_conversion">[docs]</a>    <span class="k">def</span> <span class="nf">prep_model_for_conversion</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">input_size</span><span class="p">:</span> <span class="n">Union</span><span class="p">[</span><span class="nb">tuple</span><span class="p">,</span> <span class="nb">list</span><span class="p">]</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    452. 

  452.         <span class="sd">&quot;&quot;&quot;</span>
    453. 

  453. <span class="sd">        A method for preparing the YoloV5 model for conversion to other frameworks (ONNX, CoreML etc)</span>
    454. 

  454. <span class="sd">        :param input_size: expected input size</span>
    455. 

  455. <span class="sd">        :return:</span>
    456. 

  456. <span class="sd">        &quot;&quot;&quot;</span>
    457. 

  457.         <span class="k">assert</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span><span class="p">,</span> <span class="s1">&#39;model has to be in eval mode to be converted&#39;</span>
    458. 

  458. 

  459.         <span class="c1"># Verify dummy_input from converter is of multiple of the grid size</span>
    460. 

  460.         <span class="n">max_stride</span> <span class="o">=</span> <span class="nb">int</span><span class="p">(</span><span class="nb">max</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">stride</span><span class="p">))</span>
    461. 

  461. 

  462.         <span class="c1"># Validate the image size</span>
    463. 

  463.         <span class="n">image_dims</span> <span class="o">=</span> <span class="n">input_size</span><span class="p">[</span><span class="o">-</span><span class="mi">2</span><span class="p">:]</span>    <span class="c1"># assume torch uses channels first layout</span>
    464. 

  464.         <span class="k">for</span> <span class="n">dim</span> <span class="ow">in</span> <span class="n">image_dims</span><span class="p">:</span>
    465. 

  465.             <span class="n">res_flag</span><span class="p">,</span> <span class="n">suggestion</span> <span class="o">=</span> <span class="n">check_img_size_divisibilty</span><span class="p">(</span><span class="n">dim</span><span class="p">,</span> <span class="n">max_stride</span><span class="p">)</span>
    466. 

  466.             <span class="k">if</span> <span class="ow">not</span> <span class="n">res_flag</span><span class="p">:</span>
    467. 

  467.                 <span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="sa">f</span><span class="s1">&#39;Invalid input size: </span><span class="si">{</span><span class="n">input_size</span><span class="si">}</span><span class="s1">. The input size must be multiple of max stride: &#39;</span>
    468. 

  468.                                  <span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">max_stride</span><span class="si">}</span><span class="s1">. The closest suggestions are: </span><span class="si">{</span><span class="n">suggestion</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="si">}</span><span class="s1">x</span><span class="si">{</span><span class="n">suggestion</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="si">}</span><span class="s1"> or &#39;</span>
    469. 

  469.                                  <span class="sa">f</span><span class="s1">&#39;</span><span class="si">{</span><span class="n">suggestion</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="si">}</span><span class="s1">x</span><span class="si">{</span><span class="n">suggestion</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="si">}</span><span class="s1">&#39;</span><span class="p">)</span>
    470. 

  470. 

  471.         <span class="c1"># Update the model with exportable operators</span>
    472. 

  472.         <span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">m</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">named_modules</span><span class="p">():</span>
    473. 

  473.             <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">Conv</span><span class="p">)</span> <span class="ow">and</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">act</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Hardswish</span><span class="p">):</span>
    474. 

  474.                 <span class="n">m</span><span class="o">.</span><span class="n">_non_persistent_buffers_set</span> <span class="o">=</span> <span class="nb">set</span><span class="p">()</span>  <span class="c1"># pytorch 1.6.0 compatibility</span>
    475. 

  475.                 <span class="n">m</span><span class="o">.</span><span class="n">act</span> <span class="o">=</span> <span class="n">ExportableHardswish</span><span class="p">()</span>  <span class="c1"># assign activation</span></div>
    476. 

  476. 

  477. <div class="viewcode-block" id="YoLoV5Base.get_include_attributes"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5Base.get_include_attributes">[docs]</a>    <span class="k">def</span> <span class="nf">get_include_attributes</span><span class="p">(</span><span class="bp">self</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="nb">list</span><span class="p">:</span>
    478. 

  478.         <span class="k">return</span> <span class="p">[</span><span class="s2">&quot;grid&quot;</span><span class="p">,</span> <span class="s2">&quot;anchors&quot;</span><span class="p">,</span> <span class="s2">&quot;anchors_grid&quot;</span><span class="p">]</span></div></div>
    479. 

  479. 

  480. 

  481. <div class="viewcode-block" id="Custom_YoLoV5"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.Custom_YoLoV5">[docs]</a><span class="k">class</span> <span class="nc">Custom_YoLoV5</span><span class="p">(</span><span class="n">YoLoV5Base</span><span class="p">):</span>
    482. 

  482.     <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="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">):</span>
    483. 

  483.         <span class="n">backbone</span> <span class="o">=</span> <span class="n">get_param</span><span class="p">(</span><span class="n">arch_params</span><span class="p">,</span> <span class="s1">&#39;backbone&#39;</span><span class="p">,</span> <span class="n">YoLoV5DarknetBackbone</span><span class="p">)</span>
    484. 

  484.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">backbone</span><span class="o">=</span><span class="n">backbone</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></div>
    485. 

  485. 

  486. 

  487. <div class="viewcode-block" id="YoLoV5S"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5S">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5S</span><span class="p">(</span><span class="n">YoLoV5Base</span><span class="p">):</span>
    488. 

  488.     <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="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">):</span>
    489. 

  489.         <span class="n">arch_params</span><span class="o">.</span><span class="n">depth_mult_factor</span> <span class="o">=</span> <span class="mf">0.33</span>
    490. 

  490.         <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span> <span class="o">=</span> <span class="mf">0.50</span>
    491. 

  491.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">backbone</span><span class="o">=</span><span class="n">YoLoV5DarknetBackbone</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></div>
    492. 

  492. 

  493. 

  494. <div class="viewcode-block" id="YoLoV5M"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5M">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5M</span><span class="p">(</span><span class="n">YoLoV5Base</span><span class="p">):</span>
    495. 

  495.     <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="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">):</span>
    496. 

  496.         <span class="n">arch_params</span><span class="o">.</span><span class="n">depth_mult_factor</span> <span class="o">=</span> <span class="mf">0.67</span>
    497. 

  497.         <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span> <span class="o">=</span> <span class="mf">0.75</span>
    498. 

  498.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">backbone</span><span class="o">=</span><span class="n">YoLoV5DarknetBackbone</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></div>
    499. 

  499. 

  500. 

  501. <div class="viewcode-block" id="YoLoV5L"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5L">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5L</span><span class="p">(</span><span class="n">YoLoV5Base</span><span class="p">):</span>
    502. 

  502.     <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="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">):</span>
    503. 

  503.         <span class="n">arch_params</span><span class="o">.</span><span class="n">depth_mult_factor</span> <span class="o">=</span> <span class="mf">1.0</span>
    504. 

  504.         <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span> <span class="o">=</span> <span class="mf">1.0</span>
    505. 

  505.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">backbone</span><span class="o">=</span><span class="n">YoLoV5DarknetBackbone</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></div>
    506. 

  506. 

  507. 

  508. <div class="viewcode-block" id="YoLoV5X"><a class="viewcode-back" href="../../../../super_gradients.training.models.html#super_gradients.training.models.yolov5.YoLoV5X">[docs]</a><span class="k">class</span> <span class="nc">YoLoV5X</span><span class="p">(</span><span class="n">YoLoV5Base</span><span class="p">):</span>
    509. 

  509.     <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="n">arch_params</span><span class="p">:</span> <span class="n">HpmStruct</span><span class="p">):</span>
    510. 

  510.         <span class="n">arch_params</span><span class="o">.</span><span class="n">depth_mult_factor</span> <span class="o">=</span> <span class="mf">1.33</span>
    511. 

  511.         <span class="n">arch_params</span><span class="o">.</span><span class="n">width_mult_factor</span> <span class="o">=</span> <span class="mf">1.25</span>
    512. 

  512.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">backbone</span><span class="o">=</span><span class="n">YoLoV5DarknetBackbone</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></div>
    513. 

  513. </pre></div>
    514. 

  514. 

  515.            </div>
    516. 

  516.           </div>
    517. 

  517.           <footer>
    518. 

  518. 

  519.   <hr/>
    520. 

  520. 

  521.   <div role="contentinfo">
    522. 

  522.     <p>&#169; Copyright 2021, SuperGradients team.</p>
    523. 

  523.   </div>
    524. 

  524. 

  525.   Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    526. 

  526.     <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    527. 

  527.     provided by <a href="https://readthedocs.org">Read the Docs</a>.
    528. 

  528.    
    529. 

  529. 

  530. </footer>
    531. 

  531.         </div>
    532. 

  532.       </div>
    533. 

  533.     </section>
    534. 

  534.   </div>
    535. 

  535.   <script>
    536. 

  536.       jQuery(function () {
    537. 

  537.           SphinxRtdTheme.Navigation.enable(true);
    538. 

  538.       });
    539. 

  539.   </script> 
    540. 

  540. 

  541. </body>
    542. 

  542. </html>
    543.
Prev
Next
Tip!

Press p or to see the previous file or, n or to see the next file

Comments

Write Preview
Loading...
Integrate AWS S3
Use S3 remote
Select bucket
Access key
Finish

Use AWS S3 as storage!

Browsing data directories saved to S3 is possible with DAGsHub. Let's configure your repository to easily display your data in the context of any commit!

Specify your S3 bucket

Access key (If needed)

Congratulations!

super-gradients is now integrated with AWS S3!

Delete Storage Key

Are you sure you want to delete this access key?

No
Yes
Integrate Google Cloud Storage
Use Google Storage
Select bucket
Upload key
Finish

Use Google Cloud Storage!

Browsing data directories saved to Google Cloud Storage is possible with DAGsHub. Let's configure your repository to easily display your data in the context of any commit!

Specify your Google Storage bucket

Service Account Key

Congratulations!

super-gradients is now integrated with Google Cloud Storage!

Delete Storage Key

Are you sure you want to delete this access key?

No
Yes
Integrate Azure Cloud Storage
Use Azure Storage
Select bucket
Set key
Finish

Use Azure Cloud Storage!

Browsing data directories saved to Azure Cloud Storage is possible with DAGsHub. Let's configure your repository to easily display your data in the context of any commit!

Specify your Azure Storage bucket

Access key (If needed)

Congratulations!

super-gradients is now integrated with Azure Cloud Storage!

Delete Storage Key

Are you sure you want to delete this access key?

No
Yes
Integrate S3 compatible storage
Use S3 like remote
Select bucket
Access key
Finish

Use any S3 compatible storage!

Browsing data directories saved to S3 compatible storage is possible with DAGsHub. Let's configure your repository to easily display your data in the context of any commit!

Specify your S3 bucket

Bucket name cannot be the same as the repository name. Please change one of them.

Access key (If needed)

Congratulations!

super-gradients is now integrated with your S3 compatible storage!

Delete Storage Key

Are you sure you want to delete this access key?

No
Yes