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

#396 Trainer constructor cleanup

Merged
Ghost merged 1 commits into Deci-AI:master from deci-ai:feature/SG-266_clean_trainer_ctor
Conversation 0 Files changed 49 Commits 1
File Compare
super-gradients
/
docs
/
_modules
/
super_gradients
/
training
/
utils
/
callbacks.html
View left file View right file
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885 
  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.utils.callbacks &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.       <link rel="stylesheet" href="../../../../_static/graphviz.css" type="text/css" />
    10. 

  10.   <!--[if lt IE 9]>
    11. 

  11.     <script src="../../../../_static/js/html5shiv.min.js"></script>
    12. 

  12.   <![endif]-->
    13. 

  13.   
    14. 

  14.         <script data-url_root="../../../../" id="documentation_options" src="../../../../_static/documentation_options.js"></script>
    15. 

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

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

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

  18.     <script src="../../../../_static/js/theme.js"></script>
    19. 

  19.     <link rel="index" title="Index" href="../../../../genindex.html" />
    20. 

  20.     <link rel="search" title="Search" href="../../../../search.html" /> 
    21. 

  21. </head>
    22. 

  22. 

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

  24.   <div class="wy-grid-for-nav">
    25. 

  25.     <nav data-toggle="wy-nav-shift" class="wy-nav-side">
    26. 

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

  27.         <div class="wy-side-nav-search" >
    28. 

  28.             <a href="../../../../index.html" class="icon icon-home"> SuperGradients
    29. 

  29.           </a>
    30. 

  30. <div role="search">
    31. 

  31.   <form id="rtd-search-form" class="wy-form" action="../../../../search.html" method="get">
    32. 

  32.     <input type="text" name="q" placeholder="Search docs" />
    33. 

  33.     <input type="hidden" name="check_keywords" value="yes" />
    34. 

  34.     <input type="hidden" name="area" value="default" />
    35. 

  35.   </form>
    36. 

  36. </div>
    37. 

  37.         </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
    38. 

  38.               <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
    39. 

  39. <ul>
    40. 

  40. <li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html">Fill our 4-question quick survey! We will raffle free SuperGradients swag between those who will participate -&gt; Fill Survey</a></li>
    41. 

  41. <li class="toctree-l1"><a class="reference internal" href="../../../../welcome.html#supergradients">SuperGradients</a></li>
    42. 

  42. </ul>
    43. 

  43. <p class="caption"><span class="caption-text">Technical Documentation</span></p>
    44. 

  44. <ul>
    45. 

  45. <li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.common.html">Common package</a></li>
    46. 

  46. <li class="toctree-l1"><a class="reference internal" href="../../../../super_gradients.training.html">Training package</a></li>
    47. 

  47. </ul>
    48. 

  48. <p class="caption"><span class="caption-text">User Guide</span></p>
    49. 

  49. <ul>
    50. 

  50. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html">What is SuperGradients?</a></li>
    51. 

  51. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#introducing-the-supergradients-library">Introducing the SuperGradients library</a></li>
    52. 

  52. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#installation">Installation</a></li>
    53. 

  53. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#integrating-your-training-code-complete-walkthrough">Integrating Your Training Code - Complete Walkthrough</a></li>
    54. 

  54. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#training-parameters">Training Parameters</a></li>
    55. 

  55. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#logs-and-checkpoints">Logs and Checkpoints</a></li>
    56. 

  56. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#dataset-parameters">Dataset Parameters</a></li>
    57. 

  57. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#network-architectures">Network Architectures</a></li>
    58. 

  58. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#pretrained-models">Pretrained Models</a></li>
    59. 

  59. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#how-to-reproduce-our-training-recipes">How To Reproduce Our Training Recipes</a></li>
    60. 

  60. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#professional-tools-integration">Professional Tools Integration</a></li>
    61. 

  61. <li class="toctree-l1"><a class="reference internal" href="../../../../user_guide.html#supergradients-faq">SuperGradients FAQ</a></li>
    62. 

  62. </ul>
    63. 

  63. 

  64.         </div>
    65. 

  65.       </div>
    66. 

  66.     </nav>
    67. 

  67. 

  68.     <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
    69. 

  69.           <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
    70. 

  70.           <a href="../../../../index.html">SuperGradients</a>
    71. 

  71.       </nav>
    72. 

  72. 

  73.       <div class="wy-nav-content">
    74. 

  74.         <div class="rst-content">
    75. 

  75.           <div role="navigation" aria-label="Page navigation">
    76. 

  76.   <ul class="wy-breadcrumbs">
    77. 

  77.       <li><a href="../../../../index.html" class="icon icon-home"></a> &raquo;</li>
    78. 

  78.           <li><a href="../../../index.html">Module code</a> &raquo;</li>
    79. 

  79.       <li>super_gradients.training.utils.callbacks</li>
    80. 

  80.       <li class="wy-breadcrumbs-aside">
    81. 

  81.       </li>
    82. 

  82.   </ul>
    83. 

  83.   <hr/>
    84. 

  84. </div>
    85. 

  85.           <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
    86. 

  86.            <div itemprop="articleBody">
    87. 

  87.              
    88. 

  88.   <h1>Source code for super_gradients.training.utils.callbacks</h1><div class="highlight"><pre>
    89. 

  89. <span></span><span class="kn">import</span> <span class="nn">copy</span>
    90. 

  90. <span class="kn">import</span> <span class="nn">os</span>
    91. 

  91. <span class="kn">import</span> <span class="nn">time</span>
    92. 

  92. <span class="kn">from</span> <span class="nn">enum</span> <span class="kn">import</span> <span class="n">Enum</span>
    93. 

  93. <span class="kn">import</span> <span class="nn">math</span>
    94. 

  94. <span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
    95. 

  95. <span class="kn">import</span> <span class="nn">onnx</span>
    96. 

  96. <span class="kn">import</span> <span class="nn">onnxruntime</span>
    97. 

  97. <span class="kn">import</span> <span class="nn">torch</span>
    98. 

  98. <span class="kn">import</span> <span class="nn">signal</span>
    99. 

  99. <span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">List</span>
    100. 

  100. 

  101. <span class="kn">from</span> <span class="nn">super_gradients.common.abstractions.abstract_logger</span> <span class="kn">import</span> <span class="n">get_logger</span>
    102. 

  102. <span class="kn">from</span> <span class="nn">super_gradients.training.utils.detection_utils</span> <span class="kn">import</span> <span class="n">DetectionVisualization</span><span class="p">,</span> <span class="n">DetectionPostPredictionCallback</span>
    103. 

  103. <span class="kn">from</span> <span class="nn">super_gradients.training.utils.segmentation_utils</span> <span class="kn">import</span> <span class="n">BinarySegmentationVisualization</span>
    104. 

  104. <span class="kn">import</span> <span class="nn">cv2</span>
    105. 

  105. 

  106. <span class="n">logger</span> <span class="o">=</span> <span class="n">get_logger</span><span class="p">(</span><span class="vm">__name__</span><span class="p">)</span>
    107. 

  107. 

  108. <span class="k">try</span><span class="p">:</span>
    109. 

  109.     <span class="kn">from</span> <span class="nn">deci_lab_client.client</span> <span class="kn">import</span> <span class="n">DeciPlatformClient</span>
    110. 

  110.     <span class="kn">from</span> <span class="nn">deci_lab_client.models</span> <span class="kn">import</span> <span class="n">ModelBenchmarkState</span>
    111. 

  111.     <span class="kn">from</span> <span class="nn">deci_lab_client.models.model_metadata</span> <span class="kn">import</span> <span class="n">ModelMetadata</span>
    112. 

  112. 

  113.     <span class="n">_imported_deci_lab_failure</span> <span class="o">=</span> <span class="kc">None</span>
    114. 

  114. <span class="k">except</span> <span class="p">(</span><span class="ne">ImportError</span><span class="p">,</span> <span class="ne">NameError</span><span class="p">,</span> <span class="ne">ModuleNotFoundError</span><span class="p">)</span> <span class="k">as</span> <span class="n">import_err</span><span class="p">:</span>
    115. 

  115.     <span class="n">logger</span><span class="o">.</span><span class="n">warn</span><span class="p">(</span><span class="s2">&quot;Failed to import deci_lab_client&quot;</span><span class="p">)</span>
    116. 

  116.     <span class="n">_imported_deci_lab_failure</span> <span class="o">=</span> <span class="n">import_err</span>
    117. 

  117. 

  118. 

  119. <div class="viewcode-block" id="Phase"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.Phase">[docs]</a><span class="k">class</span> <span class="nc">Phase</span><span class="p">(</span><span class="n">Enum</span><span class="p">):</span>
    120. 

  120.     <span class="n">PRE_TRAINING</span> <span class="o">=</span> <span class="s2">&quot;PRE_TRAINING&quot;</span>
    121. 

  121.     <span class="n">TRAIN_BATCH_END</span> <span class="o">=</span> <span class="s2">&quot;TRAIN_BATCH_END&quot;</span>
    122. 

  122.     <span class="n">TRAIN_BATCH_STEP</span> <span class="o">=</span> <span class="s2">&quot;TRAIN_BATCH_STEP&quot;</span>
    123. 

  123.     <span class="n">TRAIN_EPOCH_START</span> <span class="o">=</span> <span class="s2">&quot;TRAIN_EPOCH_START&quot;</span>
    124. 

  124.     <span class="n">TRAIN_EPOCH_END</span> <span class="o">=</span> <span class="s2">&quot;TRAIN_EPOCH_END&quot;</span>
    125. 

  125.     <span class="n">VALIDATION_BATCH_END</span> <span class="o">=</span> <span class="s2">&quot;VALIDATION_BATCH_END&quot;</span>
    126. 

  126.     <span class="n">VALIDATION_EPOCH_END</span> <span class="o">=</span> <span class="s2">&quot;VALIDATION_EPOCH_END&quot;</span>
    127. 

  127.     <span class="n">VALIDATION_END_BEST_EPOCH</span> <span class="o">=</span> <span class="s2">&quot;VALIDATION_END_BEST_EPOCH&quot;</span>
    128. 

  128.     <span class="n">TEST_BATCH_END</span> <span class="o">=</span> <span class="s2">&quot;TEST_BATCH_END&quot;</span>
    129. 

  129.     <span class="n">TEST_END</span> <span class="o">=</span> <span class="s2">&quot;TEST_END&quot;</span>
    130. 

  130.     <span class="n">POST_TRAINING</span> <span class="o">=</span> <span class="s2">&quot;POST_TRAINING&quot;</span></div>
    131. 

  131. 

  132. 

  133. <div class="viewcode-block" id="ContextSgMethods"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.ContextSgMethods">[docs]</a><span class="k">class</span> <span class="nc">ContextSgMethods</span><span class="p">:</span>
    134. 

  134.     <span class="sd">&quot;&quot;&quot;</span>
    135. 

  135. <span class="sd">    Class for delegating SgModel&#39;s methods, so that only the relevant ones are (&quot;phase wise&quot;) are accessible.</span>
    136. 

  136. <span class="sd">    &quot;&quot;&quot;</span>
    137. 

  137.     <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">**</span><span class="n">methods</span><span class="p">):</span>
    138. 

  138.         <span class="k">for</span> <span class="n">attr</span><span class="p">,</span> <span class="n">attr_val</span> <span class="ow">in</span> <span class="n">methods</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
    139. 

  139.             <span class="nb">setattr</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">attr</span><span class="p">,</span> <span class="n">attr_val</span><span class="p">)</span></div>
    140. 

  140. 

  141. 

  142. <div class="viewcode-block" id="PhaseContext"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PhaseContext">[docs]</a><span class="k">class</span> <span class="nc">PhaseContext</span><span class="p">:</span>
    143. 

  143.     <span class="sd">&quot;&quot;&quot;</span>
    144. 

  144. <span class="sd">    Represents the input for phase callbacks, and is constantly updated after callback calls.</span>
    145. 

  145. 

  146. <span class="sd">    &quot;&quot;&quot;</span>
    147. 

  147. 

  148.     <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">epoch</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">batch_idx</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">optimizer</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">metrics_dict</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">inputs</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">preds</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    149. 

  149.                  <span class="n">target</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">metrics_compute_fn</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">loss_avg_meter</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">loss_log_items</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">criterion</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    150. 

  150.                  <span class="n">device</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">experiment_name</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">ckpt_dir</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">net</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">lr_warmup_epochs</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">sg_logger</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    151. 

  151.                  <span class="n">train_loader</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">valid_loader</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    152. 

  152.                  <span class="n">training_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">ddp_silent_mode</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">checkpoint_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">architecture</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    153. 

  153.                  <span class="n">arch_params</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">metric_idx_in_results_tuple</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span>
    154. 

  154.                  <span class="n">metric_to_watch</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">valid_metrics</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">context_methods</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
    155. 

  155.         <span class="bp">self</span><span class="o">.</span><span class="n">epoch</span> <span class="o">=</span> <span class="n">epoch</span>
    156. 

  156.         <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span> <span class="o">=</span> <span class="n">batch_idx</span>
    157. 

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

  158.         <span class="bp">self</span><span class="o">.</span><span class="n">inputs</span> <span class="o">=</span> <span class="n">inputs</span>
    159. 

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

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

  161.         <span class="bp">self</span><span class="o">.</span><span class="n">metrics_dict</span> <span class="o">=</span> <span class="n">metrics_dict</span>
    162. 

  162.         <span class="bp">self</span><span class="o">.</span><span class="n">metrics_compute_fn</span> <span class="o">=</span> <span class="n">metrics_compute_fn</span>
    163. 

  163.         <span class="bp">self</span><span class="o">.</span><span class="n">loss_avg_meter</span> <span class="o">=</span> <span class="n">loss_avg_meter</span>
    164. 

  164.         <span class="bp">self</span><span class="o">.</span><span class="n">loss_log_items</span> <span class="o">=</span> <span class="n">loss_log_items</span>
    165. 

  165.         <span class="bp">self</span><span class="o">.</span><span class="n">criterion</span> <span class="o">=</span> <span class="n">criterion</span>
    166. 

  166.         <span class="bp">self</span><span class="o">.</span><span class="n">device</span> <span class="o">=</span> <span class="n">device</span>
    167. 

  167.         <span class="bp">self</span><span class="o">.</span><span class="n">stop_training</span> <span class="o">=</span> <span class="kc">False</span>
    168. 

  168.         <span class="bp">self</span><span class="o">.</span><span class="n">experiment_name</span> <span class="o">=</span> <span class="n">experiment_name</span>
    169. 

  169.         <span class="bp">self</span><span class="o">.</span><span class="n">ckpt_dir</span> <span class="o">=</span> <span class="n">ckpt_dir</span>
    170. 

  170.         <span class="bp">self</span><span class="o">.</span><span class="n">net</span> <span class="o">=</span> <span class="n">net</span>
    171. 

  171.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">=</span> <span class="n">lr_warmup_epochs</span>
    172. 

  172.         <span class="bp">self</span><span class="o">.</span><span class="n">sg_logger</span> <span class="o">=</span> <span class="n">sg_logger</span>
    173. 

  173.         <span class="bp">self</span><span class="o">.</span><span class="n">train_loader</span> <span class="o">=</span> <span class="n">train_loader</span>
    174. 

  174.         <span class="bp">self</span><span class="o">.</span><span class="n">valid_loader</span> <span class="o">=</span> <span class="n">valid_loader</span>
    175. 

  175.         <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span> <span class="o">=</span> <span class="n">training_params</span>
    176. 

  176.         <span class="bp">self</span><span class="o">.</span><span class="n">ddp_silent_mode</span> <span class="o">=</span> <span class="n">ddp_silent_mode</span>
    177. 

  177.         <span class="bp">self</span><span class="o">.</span><span class="n">checkpoint_params</span> <span class="o">=</span> <span class="n">checkpoint_params</span>
    178. 

  178.         <span class="bp">self</span><span class="o">.</span><span class="n">architecture</span> <span class="o">=</span> <span class="n">architecture</span>
    179. 

  179.         <span class="bp">self</span><span class="o">.</span><span class="n">arch_params</span> <span class="o">=</span> <span class="n">arch_params</span>
    180. 

  180.         <span class="bp">self</span><span class="o">.</span><span class="n">metric_idx_in_results_tuple</span> <span class="o">=</span> <span class="n">metric_idx_in_results_tuple</span>
    181. 

  181.         <span class="bp">self</span><span class="o">.</span><span class="n">metric_to_watch</span> <span class="o">=</span> <span class="n">metric_to_watch</span>
    182. 

  182.         <span class="bp">self</span><span class="o">.</span><span class="n">valid_metrics</span> <span class="o">=</span> <span class="n">valid_metrics</span>
    183. 

  183.         <span class="bp">self</span><span class="o">.</span><span class="n">context_methods</span> <span class="o">=</span> <span class="n">context_methods</span>
    184. 

  184. 

  185. <div class="viewcode-block" id="PhaseContext.update_context"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PhaseContext.update_context">[docs]</a>    <span class="k">def</span> <span class="nf">update_context</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    186. 

  186.         <span class="k">for</span> <span class="n">attr</span><span class="p">,</span> <span class="n">attr_val</span> <span class="ow">in</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
    187. 

  187.             <span class="nb">setattr</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">attr</span><span class="p">,</span> <span class="n">attr_val</span><span class="p">)</span></div></div>
    188. 

  188. 

  189. 

  190. <div class="viewcode-block" id="PhaseCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PhaseCallback">[docs]</a><span class="k">class</span> <span class="nc">PhaseCallback</span><span class="p">:</span>
    191. 

  191.     <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">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">):</span>
    192. 

  192.         <span class="bp">self</span><span class="o">.</span><span class="n">phase</span> <span class="o">=</span> <span class="n">phase</span>
    193. 

  193. 

  194.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    195. 

  195.         <span class="k">raise</span> <span class="ne">NotImplementedError</span>
    196. 

  196. 

  197.     <span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    198. 

  198.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span></div>
    199. 

  199. 

  200. 

  201. <div class="viewcode-block" id="ModelConversionCheckCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.ModelConversionCheckCallback">[docs]</a><span class="k">class</span> <span class="nc">ModelConversionCheckCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    202. 

  202.     <span class="sd">&quot;&quot;&quot;</span>
    203. 

  203. <span class="sd">    Pre-training callback that verifies model conversion to onnx given specified conversion parameters.</span>
    204. 

  204. 

  205. <span class="sd">    The model is converted, then inference is applied with onnx runtime.</span>
    206. 

  206. 

  207. <span class="sd">    Use this callback wit hthe same args as DeciPlatformCallback to prevent conversion fails at the end of training.</span>
    208. 

  208. 

  209. <span class="sd">    Attributes:</span>
    210. 

  210. 

  211. <span class="sd">        model_meta_data: (ModelMetadata) model&#39;s meta-data object.</span>
    212. 

  212. 

  213. <span class="sd">        The following parameters may be passed as kwargs in order to control the conversion to onnx:</span>
    214. 

  214. <span class="sd">        :param opset_version (default=11)</span>
    215. 

  215. <span class="sd">        :param do_constant_folding (default=True)</span>
    216. 

  216. <span class="sd">        :param dynamic_axes (default=</span>
    217. 

  217. <span class="sd">                                        {&#39;input&#39;: {0: &#39;batch_size&#39;},</span>
    218. 

  218. <span class="sd">                                        # Variable length axes</span>
    219. 

  219. <span class="sd">                                        &#39;output&#39;: {0: &#39;batch_size&#39;}}</span>
    220. 

  220. 

  221. <span class="sd">                                        )</span>
    222. 

  222. <span class="sd">        :param input_names (default=[&quot;input&quot;])</span>
    223. 

  223. <span class="sd">        :param output_names (default=[&quot;output&quot;])</span>
    224. 

  224. <span class="sd">        :param rtol (default=1e-03)</span>
    225. 

  225. <span class="sd">        :param atol (default=1e-05)</span>
    226. 

  226. <span class="sd">    &quot;&quot;&quot;</span>
    227. 

  227. 

  228.     <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">model_meta_data</span><span class="p">:</span> <span class="n">ModelMetadata</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    229. 

  229.         <span class="nb">super</span><span class="p">(</span><span class="n">ModelConversionCheckCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="o">=</span><span class="n">Phase</span><span class="o">.</span><span class="n">PRE_TRAINING</span><span class="p">)</span>
    230. 

  230.         <span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span> <span class="o">=</span> <span class="n">model_meta_data</span>
    231. 

  231. 

  232.         <span class="bp">self</span><span class="o">.</span><span class="n">opset_version</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;opset_version&quot;</span><span class="p">,</span> <span class="mi">10</span><span class="p">)</span>
    233. 

  233.         <span class="bp">self</span><span class="o">.</span><span class="n">do_constant_folding</span> <span class="o">=</span> <span class="p">(</span>
    234. 

  234.             <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;do_constant_folding&quot;</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span> <span class="k">if</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;do_constant_folding&quot;</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span> <span class="k">else</span> <span class="kc">True</span>
    235. 

  235.         <span class="p">)</span>
    236. 

  236.         <span class="bp">self</span><span class="o">.</span><span class="n">input_names</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;input_names&quot;</span><span class="p">)</span> <span class="ow">or</span> <span class="p">[</span><span class="s2">&quot;input&quot;</span><span class="p">]</span>
    237. 

  237.         <span class="bp">self</span><span class="o">.</span><span class="n">output_names</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;output_names&quot;</span><span class="p">)</span> <span class="ow">or</span> <span class="p">[</span><span class="s2">&quot;output&quot;</span><span class="p">]</span>
    238. 

  238.         <span class="bp">self</span><span class="o">.</span><span class="n">dynamic_axes</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;dynamic_axes&quot;</span><span class="p">)</span> <span class="ow">or</span> <span class="p">{</span><span class="s2">&quot;input&quot;</span><span class="p">:</span> <span class="p">{</span><span class="mi">0</span><span class="p">:</span> <span class="s2">&quot;batch_size&quot;</span><span class="p">},</span> <span class="s2">&quot;output&quot;</span><span class="p">:</span> <span class="p">{</span><span class="mi">0</span><span class="p">:</span> <span class="s2">&quot;batch_size&quot;</span><span class="p">}}</span>
    239. 

  239. 

  240.         <span class="bp">self</span><span class="o">.</span><span class="n">rtol</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;rtol&quot;</span><span class="p">,</span> <span class="mf">1e-03</span><span class="p">)</span>
    241. 

  241.         <span class="bp">self</span><span class="o">.</span><span class="n">atol</span> <span class="o">=</span> <span class="n">kwargs</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s2">&quot;atol&quot;</span><span class="p">,</span> <span class="mf">1e-05</span><span class="p">)</span>
    242. 

  242. 

  243.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    244. 

  244.         <span class="n">model</span> <span class="o">=</span> <span class="n">copy</span><span class="o">.</span><span class="n">deepcopy</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">net</span><span class="o">.</span><span class="n">module</span><span class="p">)</span>
    245. 

  245.         <span class="n">model</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span>
    246. 

  246.         <span class="n">model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>  <span class="c1"># Put model into eval mode</span>
    247. 

  247. 

  248.         <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="s2">&quot;prep_model_for_conversion&quot;</span><span class="p">):</span>
    249. 

  249.             <span class="n">model</span><span class="o">.</span><span class="n">prep_model_for_conversion</span><span class="p">(</span><span class="n">input_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">input_dimensions</span><span class="p">)</span>
    250. 

  250. 

  251.         <span class="n">x</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span>
    252. 

  252.             <span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">primary_batch_size</span><span class="p">,</span> <span class="o">*</span><span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">input_dimensions</span><span class="p">,</span> <span class="n">requires_grad</span><span class="o">=</span><span class="kc">False</span>
    253. 

  253.         <span class="p">)</span>
    254. 

  254. 

  255.         <span class="n">tmp_model_path</span> <span class="o">=</span> <span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">ckpt_dir</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">name</span> <span class="o">+</span> <span class="s2">&quot;_tmp.onnx&quot;</span><span class="p">)</span>
    256. 

  256. 

  257.         <span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span>
    258. 

  258.             <span class="n">torch_out</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
    259. 

  259. 

  260.         <span class="n">torch</span><span class="o">.</span><span class="n">onnx</span><span class="o">.</span><span class="n">export</span><span class="p">(</span>
    261. 

  261.             <span class="n">model</span><span class="p">,</span>  <span class="c1"># Model being run</span>
    262. 

  262.             <span class="n">x</span><span class="p">,</span>  <span class="c1"># Model input (or a tuple for multiple inputs)</span>
    263. 

  263.             <span class="n">tmp_model_path</span><span class="p">,</span>  <span class="c1"># Where to save the model (can be a file or file-like object)</span>
    264. 

  264.             <span class="n">export_params</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>  <span class="c1"># Store the trained parameter weights inside the model file</span>
    265. 

  265.             <span class="n">opset_version</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">opset_version</span><span class="p">,</span>
    266. 

  266.             <span class="n">do_constant_folding</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">do_constant_folding</span><span class="p">,</span>
    267. 

  267.             <span class="n">input_names</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">input_names</span><span class="p">,</span>
    268. 

  268.             <span class="n">output_names</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">output_names</span><span class="p">,</span>
    269. 

  269.             <span class="n">dynamic_axes</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">dynamic_axes</span><span class="p">,</span>
    270. 

  270.         <span class="p">)</span>
    271. 

  271. 

  272.         <span class="n">onnx_model</span> <span class="o">=</span> <span class="n">onnx</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">tmp_model_path</span><span class="p">)</span>
    273. 

  273.         <span class="n">onnx</span><span class="o">.</span><span class="n">checker</span><span class="o">.</span><span class="n">check_model</span><span class="p">(</span><span class="n">onnx_model</span><span class="p">)</span>
    274. 

  274. 

  275.         <span class="n">ort_session</span> <span class="o">=</span> <span class="n">onnxruntime</span><span class="o">.</span><span class="n">InferenceSession</span><span class="p">(</span>
    276. 

  276.             <span class="n">tmp_model_path</span><span class="p">,</span> <span class="n">providers</span><span class="o">=</span><span class="p">[</span><span class="s2">&quot;CUDAExecutionProvider&quot;</span><span class="p">,</span> <span class="s2">&quot;CPUExecutionProvider&quot;</span><span class="p">]</span>
    277. 

  277.         <span class="p">)</span>
    278. 

  278. 

  279.         <span class="c1"># compute ONNX Runtime output prediction</span>
    280. 

  280.         <span class="n">ort_inputs</span> <span class="o">=</span> <span class="p">{</span><span class="n">ort_session</span><span class="o">.</span><span class="n">get_inputs</span><span class="p">()[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">name</span><span class="p">:</span> <span class="n">x</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">numpy</span><span class="p">()}</span>
    281. 

  281.         <span class="n">ort_outs</span> <span class="o">=</span> <span class="n">ort_session</span><span class="o">.</span><span class="n">run</span><span class="p">(</span><span class="kc">None</span><span class="p">,</span> <span class="n">ort_inputs</span><span class="p">)</span>
    282. 

  282. 

  283.         <span class="c1"># TODO: Ideally we don&#39;t want to check this but have the certainty of just calling torch_out.cpu()</span>
    284. 

  284.         <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">torch_out</span><span class="p">,</span> <span class="n">List</span><span class="p">)</span> <span class="ow">or</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">torch_out</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">):</span>
    285. 

  285.             <span class="n">torch_out</span> <span class="o">=</span> <span class="n">torch_out</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span>
    286. 

  286.         <span class="c1"># compare ONNX Runtime and PyTorch results</span>
    287. 

  287.         <span class="n">np</span><span class="o">.</span><span class="n">testing</span><span class="o">.</span><span class="n">assert_allclose</span><span class="p">(</span><span class="n">torch_out</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span><span class="o">.</span><span class="n">numpy</span><span class="p">(),</span> <span class="n">ort_outs</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">rtol</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">rtol</span><span class="p">,</span> <span class="n">atol</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">atol</span><span class="p">)</span>
    288. 

  288. 

  289.         <span class="n">os</span><span class="o">.</span><span class="n">remove</span><span class="p">(</span><span class="n">tmp_model_path</span><span class="p">)</span>
    290. 

  290. 

  291.         <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">&quot;Exported model has been tested with ONNXRuntime, and the result looks good!&quot;</span><span class="p">)</span></div>
    292. 

  292. 

  293. 

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

  295.     <span class="sd">&quot;&quot;&quot;</span>
    296. 

  296. <span class="sd">    Post-training callback for uploading and optimizing a model.</span>
    297. 

  297. 

  298. <span class="sd">    Attributes:</span>
    299. 

  299. 

  300. <span class="sd">        email: (str) username for Deci platform.</span>
    301. 

  301. <span class="sd">        model_meta_data: (ModelMetadata) model&#39;s meta-data object.</span>
    302. 

  302. <span class="sd">        optimization_request_form: (dict) optimization request form object.</span>
    303. 

  303. <span class="sd">        password: (str) default=None, should only be used for testing.</span>
    304. 

  304. <span class="sd">        ckpt_name: (str) default=&quot;ckpt_best&quot; refers to the filename of the checkpoint, inside the checkpoint directory.</span>
    305. 

  305. 

  306. <span class="sd">        The following parameters may be passed as kwargs in order to control the conversion to onnx:</span>
    307. 

  307. <span class="sd">        :param opset_version</span>
    308. 

  308. <span class="sd">        :param do_constant_folding</span>
    309. 

  309. <span class="sd">        :param dynamic_axes</span>
    310. 

  310. <span class="sd">        :param input_names</span>
    311. 

  311. <span class="sd">        :param output_names</span>
    312. 

  312. <span class="sd">    &quot;&quot;&quot;</span>
    313. 

  313. 

  314.     <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">model_meta_data</span><span class="p">,</span> <span class="n">optimization_request_form</span><span class="p">,</span> <span class="n">auth_token</span><span class="p">:</span> <span class="nb">str</span> <span class="o">=</span> <span class="kc">None</span><span class="p">,</span> <span class="n">ckpt_name</span><span class="o">=</span><span class="s2">&quot;ckpt_best.pth&quot;</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    315. 

  315.         <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">phase</span><span class="o">=</span><span class="n">Phase</span><span class="o">.</span><span class="n">POST_TRAINING</span><span class="p">)</span>
    316. 

  316.         <span class="k">if</span> <span class="n">_imported_deci_lab_failure</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
    317. 

  317.             <span class="k">raise</span> <span class="n">_imported_deci_lab_failure</span>
    318. 

  318.         <span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span> <span class="o">=</span> <span class="n">model_meta_data</span>
    319. 

  319.         <span class="bp">self</span><span class="o">.</span><span class="n">optimization_request_form</span> <span class="o">=</span> <span class="n">optimization_request_form</span>
    320. 

  320.         <span class="bp">self</span><span class="o">.</span><span class="n">conversion_kwargs</span> <span class="o">=</span> <span class="n">kwargs</span>
    321. 

  321.         <span class="bp">self</span><span class="o">.</span><span class="n">ckpt_name</span> <span class="o">=</span> <span class="n">ckpt_name</span>
    322. 

  322.         <span class="bp">self</span><span class="o">.</span><span class="n">platform_client</span> <span class="o">=</span> <span class="n">DeciPlatformClient</span><span class="p">(</span><span class="s2">&quot;api.deci.ai&quot;</span><span class="p">,</span> <span class="mi">443</span><span class="p">,</span> <span class="n">https</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
    323. 

  323.         <span class="bp">self</span><span class="o">.</span><span class="n">platform_client</span><span class="o">.</span><span class="n">login</span><span class="p">(</span><span class="n">token</span><span class="o">=</span><span class="n">auth_token</span><span class="p">)</span>
    324. 

  324. 

  325. <div class="viewcode-block" id="DeciLabUploadCallback.log_optimization_failed"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.DeciLabUploadCallback.log_optimization_failed">[docs]</a>    <span class="nd">@staticmethod</span>
    326. 

  326.     <span class="k">def</span> <span class="nf">log_optimization_failed</span><span class="p">():</span>
    327. 

  327.         <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">&quot;We couldn&#39;t finish your model optimization. Visit https://console.deci.ai for details&quot;</span><span class="p">)</span></div>
    328. 

  328. 

  329. <div class="viewcode-block" id="DeciLabUploadCallback.upload_model"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.DeciLabUploadCallback.upload_model">[docs]</a>    <span class="k">def</span> <span class="nf">upload_model</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">model</span><span class="p">):</span>
    330. 

  330.         <span class="sd">&quot;&quot;&quot;</span>
    331. 

  331. <span class="sd">            This function will upload the trained model to the Deci Lab</span>
    332. 

  332. 

  333. <span class="sd">            Args:</span>
    334. 

  334. <span class="sd">                model: The resulting model from the training process</span>
    335. 

  335. <span class="sd">        &quot;&quot;&quot;</span>
    336. 

  336.         <span class="bp">self</span><span class="o">.</span><span class="n">platform_client</span><span class="o">.</span><span class="n">add_model</span><span class="p">(</span>
    337. 

  337.             <span class="n">add_model_request</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="p">,</span>
    338. 

  338.             <span class="n">optimization_request</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">optimization_request_form</span><span class="p">,</span>
    339. 

  339.             <span class="n">local_loaded_model</span><span class="o">=</span><span class="n">model</span><span class="p">,</span>
    340. 

  340.         <span class="p">)</span></div>
    341. 

  341. 

  342. <div class="viewcode-block" id="DeciLabUploadCallback.get_optimization_status"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.DeciLabUploadCallback.get_optimization_status">[docs]</a>    <span class="k">def</span> <span class="nf">get_optimization_status</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">optimized_model_name</span><span class="p">:</span> <span class="nb">str</span><span class="p">):</span>
    343. 

  343.         <span class="sd">&quot;&quot;&quot;</span>
    344. 

  344. <span class="sd">            This function will do fetch the optimized version of the trained model and check on its benchmark status.</span>
    345. 

  345. <span class="sd">            The status will be checked against the server every 30 seconds and the process will timeout after 30 minutes</span>
    346. 

  346. <span class="sd">            or log about the successful optimization - whichever happens first.</span>
    347. 

  347. <span class="sd">            Args:</span>
    348. 

  348. <span class="sd">                optimized_model_name (str): Optimized model name</span>
    349. 

  349. <span class="sd">            Returns:</span>
    350. 

  350. <span class="sd">                bool: whether or not the optimized model has been benchmarked</span>
    351. 

  351. <span class="sd">        &quot;&quot;&quot;</span>
    352. 

  352. 

  353.         <span class="k">def</span> <span class="nf">handler</span><span class="p">(</span><span class="n">_signum</span><span class="p">,</span> <span class="n">_frame</span><span class="p">):</span>
    354. 

  354.             <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="s2">&quot;Process timed out. Visit https://console.deci.ai for details&quot;</span><span class="p">)</span>
    355. 

  355.             <span class="k">return</span> <span class="kc">False</span>
    356. 

  356. 

  357.         <span class="n">signal</span><span class="o">.</span><span class="n">signal</span><span class="p">(</span><span class="n">signal</span><span class="o">.</span><span class="n">SIGALRM</span><span class="p">,</span> <span class="n">handler</span><span class="p">)</span>
    358. 

  358.         <span class="n">signal</span><span class="o">.</span><span class="n">alarm</span><span class="p">(</span><span class="mi">1800</span><span class="p">)</span>
    359. 

  359. 

  360.         <span class="n">finished</span> <span class="o">=</span> <span class="kc">False</span>
    361. 

  361.         <span class="k">while</span> <span class="ow">not</span> <span class="n">finished</span><span class="p">:</span>
    362. 

  362.             <span class="n">optimized_model</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">platform_client</span><span class="o">.</span><span class="n">get_model_by_name</span><span class="p">(</span><span class="n">name</span><span class="o">=</span><span class="n">optimized_model_name</span><span class="p">)</span><span class="o">.</span><span class="n">data</span>
    363. 

  363.             <span class="k">if</span> <span class="n">optimized_model</span><span class="o">.</span><span class="n">benchmark_state</span> <span class="ow">not</span> <span class="ow">in</span> <span class="p">[</span><span class="n">ModelBenchmarkState</span><span class="o">.</span><span class="n">IN_PROGRESS</span><span class="p">,</span> <span class="n">ModelBenchmarkState</span><span class="o">.</span><span class="n">PENDING</span><span class="p">]:</span>
    364. 

  364.                 <span class="n">finished</span> <span class="o">=</span> <span class="kc">True</span>
    365. 

  365.             <span class="k">else</span><span class="p">:</span>
    366. 

  366.                 <span class="n">time</span><span class="o">.</span><span class="n">sleep</span><span class="p">(</span><span class="mi">30</span><span class="p">)</span>
    367. 

  367. 

  368.         <span class="n">signal</span><span class="o">.</span><span class="n">alarm</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
    369. 

  369.         <span class="k">return</span> <span class="kc">True</span></div>
    370. 

  370. 

  371.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    372. 

  372.         <span class="sd">&quot;&quot;&quot;</span>
    373. 

  373. <span class="sd">            This function will attempt to upload the trained model and schedule an optimization for it.</span>
    374. 

  374. <span class="sd">            Args:</span>
    375. 

  375. <span class="sd">                context (PhaseContext): Training phase context</span>
    376. 

  376. <span class="sd">            Returns:</span>
    377. 

  377. <span class="sd">                bool: whether or not the optimized model has been benchmarked</span>
    378. 

  378. <span class="sd">            &quot;&quot;&quot;</span>
    379. 

  379.         <span class="k">try</span><span class="p">:</span>
    380. 

  380.             <span class="n">model</span> <span class="o">=</span> <span class="n">copy</span><span class="o">.</span><span class="n">deepcopy</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">net</span><span class="p">)</span>
    381. 

  381.             <span class="n">model_state_dict_path</span> <span class="o">=</span> <span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">ckpt_dir</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">ckpt_name</span><span class="p">)</span>
    382. 

  382.             <span class="n">model_state_dict</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">model_state_dict_path</span><span class="p">)[</span><span class="s2">&quot;net&quot;</span><span class="p">]</span>
    383. 

  383.             <span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">state_dict</span><span class="o">=</span><span class="n">model_state_dict</span><span class="p">)</span>
    384. 

  384. 

  385.             <span class="n">model</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span>
    386. 

  386.             <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="s2">&quot;prep_model_for_conversion&quot;</span><span class="p">):</span>
    387. 

  387.                 <span class="n">model</span><span class="o">.</span><span class="n">prep_model_for_conversion</span><span class="p">(</span><span class="n">input_size</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">input_dimensions</span><span class="p">)</span>
    388. 

  388. 

  389.             <span class="bp">self</span><span class="o">.</span><span class="n">upload_model</span><span class="p">(</span><span class="n">model</span><span class="o">=</span><span class="n">model</span><span class="p">)</span>
    390. 

  390.             <span class="n">model_name</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">model_meta_data</span><span class="o">.</span><span class="n">name</span>
    391. 

  391.             <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="sa">f</span><span class="s2">&quot;Successfully added </span><span class="si">{</span><span class="n">model_name</span><span class="si">}</span><span class="s2"> to the model repository&quot;</span><span class="p">)</span>
    392. 

  392. 

  393.             <span class="n">optimized_model_name</span> <span class="o">=</span> <span class="sa">f</span><span class="s2">&quot;</span><span class="si">{</span><span class="n">model_name</span><span class="si">}</span><span class="s2">_1_1&quot;</span>
    394. 

  394.             <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">&quot;We&#39;ll wait for the scheduled optimization to finish. Please don&#39;t close this window&quot;</span><span class="p">)</span>
    395. 

  395.             <span class="n">success</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">get_optimization_status</span><span class="p">(</span><span class="n">optimized_model_name</span><span class="o">=</span><span class="n">optimized_model_name</span><span class="p">)</span>
    396. 

  396.             <span class="k">if</span> <span class="n">success</span><span class="p">:</span>
    397. 

  397.                 <span class="n">logger</span><span class="o">.</span><span class="n">info</span><span class="p">(</span><span class="s2">&quot;Successfully finished your model optimization. Visit https://console.deci.ai for details&quot;</span><span class="p">)</span>
    398. 

  398.             <span class="k">else</span><span class="p">:</span>
    399. 

  399.                 <span class="n">DeciLabUploadCallback</span><span class="o">.</span><span class="n">log_optimization_failed</span><span class="p">()</span>
    400. 

  400.         <span class="k">except</span> <span class="ne">Exception</span> <span class="k">as</span> <span class="n">ex</span><span class="p">:</span>
    401. 

  401.             <span class="n">DeciLabUploadCallback</span><span class="o">.</span><span class="n">log_optimization_failed</span><span class="p">()</span>
    402. 

  402.             <span class="n">logger</span><span class="o">.</span><span class="n">error</span><span class="p">(</span><span class="n">ex</span><span class="p">)</span></div>
    403. 

  403. 

  404. 

  405. <div class="viewcode-block" id="LRCallbackBase"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.LRCallbackBase">[docs]</a><span class="k">class</span> <span class="nc">LRCallbackBase</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    406. 

  406.     <span class="sd">&quot;&quot;&quot;</span>
    407. 

  407. <span class="sd">    Base class for hard coded learning rate scheduling regimes, implemented as callbacks.</span>
    408. 

  408. <span class="sd">    &quot;&quot;&quot;</span>
    409. 

  409. 

  410.     <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">phase</span><span class="p">,</span> <span class="n">initial_lr</span><span class="p">,</span> <span class="n">update_param_groups</span><span class="p">,</span> <span class="n">train_loader_len</span><span class="p">,</span> <span class="n">net</span><span class="p">,</span> <span class="n">training_params</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    411. 

  411.         <span class="nb">super</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    412. 

  412.         <span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">=</span> <span class="n">initial_lr</span>
    413. 

  413.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="n">initial_lr</span>
    414. 

  414.         <span class="bp">self</span><span class="o">.</span><span class="n">update_param_groups</span> <span class="o">=</span> <span class="n">update_param_groups</span>
    415. 

  415.         <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">=</span> <span class="n">train_loader_len</span>
    416. 

  416.         <span class="bp">self</span><span class="o">.</span><span class="n">net</span> <span class="o">=</span> <span class="n">net</span>
    417. 

  417.         <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span> <span class="o">=</span> <span class="n">training_params</span>
    418. 

  418. 

  419.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    420. 

  420.         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">is_lr_scheduling_enabled</span><span class="p">(</span><span class="n">context</span><span class="p">):</span>
    421. 

  421.             <span class="bp">self</span><span class="o">.</span><span class="n">perform_scheduling</span><span class="p">(</span><span class="n">context</span><span class="p">)</span>
    422. 

  422. 

  423. <div class="viewcode-block" id="LRCallbackBase.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.LRCallbackBase.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    424. 

  424.         <span class="sd">&quot;&quot;&quot;</span>
    425. 

  425. <span class="sd">        Predicate that controls whether to perform lr scheduling based on values in context.</span>
    426. 

  426. 

  427. <span class="sd">        @param context: PhaseContext: current phase&#39;s context.</span>
    428. 

  428. <span class="sd">        @return: bool, whether to apply lr scheduling or not.</span>
    429. 

  429. <span class="sd">        &quot;&quot;&quot;</span>
    430. 

  430.         <span class="k">raise</span> <span class="ne">NotImplementedError</span></div>
    431. 

  431. 

  432. <div class="viewcode-block" id="LRCallbackBase.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.LRCallbackBase.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    433. 

  433.         <span class="sd">&quot;&quot;&quot;</span>
    434. 

  434. <span class="sd">        Performs lr scheduling based on values in context.</span>
    435. 

  435. 

  436. <span class="sd">        @param context: PhaseContext: current phase&#39;s context.</span>
    437. 

  437. <span class="sd">        &quot;&quot;&quot;</span>
    438. 

  438.         <span class="k">raise</span> <span class="ne">NotImplementedError</span></div>
    439. 

  439. 

  440. <div class="viewcode-block" id="LRCallbackBase.update_lr"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.LRCallbackBase.update_lr">[docs]</a>    <span class="k">def</span> <span class="nf">update_lr</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">optimizer</span><span class="p">,</span> <span class="n">epoch</span><span class="p">,</span> <span class="n">batch_idx</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
    441. 

  441.         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">update_param_groups</span><span class="p">:</span>
    442. 

  442.             <span class="n">param_groups</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">net</span><span class="o">.</span><span class="n">module</span><span class="o">.</span><span class="n">update_param_groups</span><span class="p">(</span>
    443. 

  443.                 <span class="n">optimizer</span><span class="o">.</span><span class="n">param_groups</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">lr</span><span class="p">,</span> <span class="n">epoch</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span>
    444. 

  444.             <span class="p">)</span>
    445. 

  445.             <span class="n">optimizer</span><span class="o">.</span><span class="n">param_groups</span> <span class="o">=</span> <span class="n">param_groups</span>
    446. 

  446.         <span class="k">else</span><span class="p">:</span>
    447. 

  447.             <span class="c1"># UPDATE THE OPTIMIZERS PARAMETER</span>
    448. 

  448.             <span class="k">for</span> <span class="n">param_group</span> <span class="ow">in</span> <span class="n">optimizer</span><span class="o">.</span><span class="n">param_groups</span><span class="p">:</span>
    449. 

  449.                 <span class="n">param_group</span><span class="p">[</span><span class="s2">&quot;lr&quot;</span><span class="p">]</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">lr</span></div></div>
    450. 

  450. 

  451. 

  452. <div class="viewcode-block" id="WarmupLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.WarmupLRCallback">[docs]</a><span class="k">class</span> <span class="nc">WarmupLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    453. 

  453.     <span class="sd">&quot;&quot;&quot;</span>
    454. 

  454. <span class="sd">    LR scheduling callback for linear step warmup.</span>
    455. 

  455. <span class="sd">    LR climbs from warmup_initial_lr with even steps to initial lr. When warmup_initial_lr is None- LR climb starts from</span>
    456. 

  456. <span class="sd">     initial_lr/(1+warmup_epochs).</span>
    457. 

  457. 

  458. <span class="sd">    &quot;&quot;&quot;</span>
    459. 

  459. 

  460.     <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    461. 

  461.         <span class="nb">super</span><span class="p">(</span><span class="n">WarmupLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_EPOCH_START</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    462. 

  462.         <span class="bp">self</span><span class="o">.</span><span class="n">warmup_initial_lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">warmup_initial_lr</span> <span class="ow">or</span> <span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">/</span> <span class="p">(</span>
    463. 

  463.             <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">+</span> <span class="mi">1</span>
    464. 

  464.         <span class="p">)</span>
    465. 

  465.         <span class="bp">self</span><span class="o">.</span><span class="n">warmup_step_size</span> <span class="o">=</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">warmup_initial_lr</span><span class="p">)</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    466. 

  466. 

  467. <div class="viewcode-block" id="WarmupLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.WarmupLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    468. 

  468.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">warmup_initial_lr</span> <span class="o">+</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">warmup_step_size</span>
    469. 

  469.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span></div>
    470. 

  470. 

  471. <div class="viewcode-block" id="WarmupLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.WarmupLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    472. 

  472.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&gt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span></div></div>
    473. 

  473. 

  474. 

  475. <div class="viewcode-block" id="StepLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.StepLRCallback">[docs]</a><span class="k">class</span> <span class="nc">StepLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    476. 

  476.     <span class="sd">&quot;&quot;&quot;</span>
    477. 

  477. <span class="sd">    Hard coded step learning rate scheduling (i.e at specific milestones).</span>
    478. 

  478. <span class="sd">    &quot;&quot;&quot;</span>
    479. 

  479. 

  480.     <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">lr_updates</span><span class="p">,</span> <span class="n">lr_decay_factor</span><span class="p">,</span> <span class="n">step_lr_update_freq</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>
    481. 

  481.         <span class="nb">super</span><span class="p">(</span><span class="n">StepLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_EPOCH_END</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    482. 

  482.         <span class="k">if</span> <span class="n">step_lr_update_freq</span> <span class="ow">and</span> <span class="nb">len</span><span class="p">(</span><span class="n">lr_updates</span><span class="p">):</span>
    483. 

  483.             <span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span>
    484. 

  484.                 <span class="s2">&quot;Only one of [lr_updates, step_lr_update_freq] should be passed to StepLRCallback constructor&quot;</span>
    485. 

  485.             <span class="p">)</span>
    486. 

  486. 

  487.         <span class="k">if</span> <span class="n">step_lr_update_freq</span><span class="p">:</span>
    488. 

  488.             <span class="n">max_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    489. 

  489.             <span class="n">warmup_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    490. 

  490.             <span class="n">lr_updates</span> <span class="o">=</span> <span class="p">[</span>
    491. 

  491.                 <span class="nb">int</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">step_lr_update_freq</span> <span class="o">*</span> <span class="n">x</span><span class="p">))</span>
    492. 

  492.                 <span class="k">for</span> <span class="n">x</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">max_epochs</span><span class="p">)</span>
    493. 

  493.                 <span class="k">if</span> <span class="n">warmup_epochs</span> <span class="o">&lt;=</span> <span class="nb">int</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">ceil</span><span class="p">(</span><span class="n">step_lr_update_freq</span> <span class="o">*</span> <span class="n">x</span><span class="p">))</span> <span class="o">&lt;</span> <span class="n">max_epochs</span>
    494. 

  494.             <span class="p">]</span>
    495. 

  495.         <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span> <span class="o">&gt;</span> <span class="mi">0</span><span class="p">:</span>
    496. 

  496.             <span class="n">logger</span><span class="o">.</span><span class="n">warning</span><span class="p">(</span>
    497. 

  497.                 <span class="s2">&quot;Specific lr_updates were passed along with cooldown_epochs &gt; 0,&quot;</span> <span class="s2">&quot; cooldown will have no effect.&quot;</span>
    498. 

  498.             <span class="p">)</span>
    499. 

  499.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_updates</span> <span class="o">=</span> <span class="n">lr_updates</span>
    500. 

  500.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_decay_factor</span> <span class="o">=</span> <span class="n">lr_decay_factor</span>
    501. 

  501. 

  502. <div class="viewcode-block" id="StepLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.StepLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    503. 

  503.         <span class="n">num_updates_passed</span> <span class="o">=</span> <span class="p">[</span><span class="n">x</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">lr_updates</span> <span class="k">if</span> <span class="n">x</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">]</span>
    504. 

  504.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">lr_decay_factor</span> <span class="o">**</span> <span class="nb">len</span><span class="p">(</span><span class="n">num_updates_passed</span><span class="p">)</span>
    505. 

  505.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span></div>
    506. 

  506. 

  507. <div class="viewcode-block" id="StepLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.StepLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    508. 

  508.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span></div></div>
    509. 

  509. 

  510. 

  511. <div class="viewcode-block" id="ExponentialLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.ExponentialLRCallback">[docs]</a><span class="k">class</span> <span class="nc">ExponentialLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    512. 

  512.     <span class="sd">&quot;&quot;&quot;</span>
    513. 

  513. <span class="sd">    Exponential decay learning rate scheduling. Decays the learning rate by `lr_decay_factor` every epoch.</span>
    514. 

  514. <span class="sd">    &quot;&quot;&quot;</span>
    515. 

  515. 

  516.     <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">lr_decay_factor</span><span class="p">:</span> <span class="nb">float</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    517. 

  517.         <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">phase</span><span class="o">=</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_BATCH_STEP</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    518. 

  518.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_decay_factor</span> <span class="o">=</span> <span class="n">lr_decay_factor</span>
    519. 

  519. 

  520. <div class="viewcode-block" id="ExponentialLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.ExponentialLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    521. 

  521.         <span class="n">effective_epoch</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    522. 

  522.         <span class="n">current_iter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">*</span> <span class="n">effective_epoch</span> <span class="o">+</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span>
    523. 

  523.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">lr_decay_factor</span> <span class="o">**</span> <span class="p">(</span><span class="n">current_iter</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span><span class="p">)</span>
    524. 

  524.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span></div>
    525. 

  525. 

  526. <div class="viewcode-block" id="ExponentialLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.ExponentialLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    527. 

  527.         <span class="n">post_warmup_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    528. 

  528.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">&lt;</span> <span class="n">post_warmup_epochs</span></div></div>
    529. 

  529. 

  530. 

  531. <div class="viewcode-block" id="PolyLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PolyLRCallback">[docs]</a><span class="k">class</span> <span class="nc">PolyLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    532. 

  532.     <span class="sd">&quot;&quot;&quot;</span>
    533. 

  533. <span class="sd">    Hard coded polynomial decay learning rate scheduling (i.e at specific milestones).</span>
    534. 

  534. <span class="sd">    &quot;&quot;&quot;</span>
    535. 

  535. 

  536.     <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">max_epochs</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    537. 

  537.         <span class="nb">super</span><span class="p">(</span><span class="n">PolyLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_BATCH_STEP</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    538. 

  538.         <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">=</span> <span class="n">max_epochs</span>
    539. 

  539. 

  540. <div class="viewcode-block" id="PolyLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PolyLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    541. 

  541.         <span class="n">effective_epoch</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    542. 

  542.         <span class="n">effective_max_epochs</span> <span class="o">=</span> <span class="p">(</span>
    543. 

  543.             <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    544. 

  544.         <span class="p">)</span>
    545. 

  545.         <span class="n">current_iter</span> <span class="o">=</span> <span class="p">(</span>
    546. 

  546.             <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">*</span> <span class="n">effective_epoch</span> <span class="o">+</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span>
    547. 

  547.         <span class="p">)</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">batch_accumulate</span>
    548. 

  548.         <span class="n">max_iter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">*</span> <span class="n">effective_max_epochs</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">batch_accumulate</span>
    549. 

  549.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">*</span> <span class="nb">pow</span><span class="p">((</span><span class="mf">1.0</span> <span class="o">-</span> <span class="p">(</span><span class="n">current_iter</span> <span class="o">/</span> <span class="n">max_iter</span><span class="p">)),</span> <span class="mf">0.9</span><span class="p">)</span>
    550. 

  550.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span></div>
    551. 

  551. 

  552. <div class="viewcode-block" id="PolyLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PolyLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    553. 

  553.         <span class="n">post_warmup_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    554. 

  554.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">&lt;</span> <span class="n">post_warmup_epochs</span></div></div>
    555. 

  555. 

  556. 

  557. <div class="viewcode-block" id="CosineLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.CosineLRCallback">[docs]</a><span class="k">class</span> <span class="nc">CosineLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    558. 

  558.     <span class="sd">&quot;&quot;&quot;</span>
    559. 

  559. <span class="sd">    Hard coded step Cosine anealing learning rate scheduling.</span>
    560. 

  560. <span class="sd">    &quot;&quot;&quot;</span>
    561. 

  561. 

  562.     <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">max_epochs</span><span class="p">,</span> <span class="n">cosine_final_lr_ratio</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    563. 

  563.         <span class="nb">super</span><span class="p">(</span><span class="n">CosineLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_BATCH_STEP</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    564. 

  564.         <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">=</span> <span class="n">max_epochs</span>
    565. 

  565.         <span class="bp">self</span><span class="o">.</span><span class="n">cosine_final_lr_ratio</span> <span class="o">=</span> <span class="n">cosine_final_lr_ratio</span>
    566. 

  566. 

  567. <div class="viewcode-block" id="CosineLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.CosineLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    568. 

  568.         <span class="n">effective_epoch</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    569. 

  569.         <span class="n">effective_max_epochs</span> <span class="o">=</span> <span class="p">(</span>
    570. 

  570.             <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    571. 

  571.         <span class="p">)</span>
    572. 

  572.         <span class="n">current_iter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">*</span> <span class="n">effective_epoch</span> <span class="o">+</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span>
    573. 

  573.         <span class="n">max_iter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span> <span class="o">*</span> <span class="n">effective_max_epochs</span>
    574. 

  574.         <span class="n">lr</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">initial_lr</span> <span class="o">*</span> <span class="p">(</span><span class="mf">1.0</span> <span class="o">+</span> <span class="n">math</span><span class="o">.</span><span class="n">cos</span><span class="p">(</span><span class="n">current_iter</span> <span class="o">/</span> <span class="p">(</span><span class="n">max_iter</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)</span> <span class="o">*</span> <span class="n">math</span><span class="o">.</span><span class="n">pi</span><span class="p">))</span>
    575. 

  575.         <span class="c1"># the cosine starts from initial_lr and reaches initial_lr * cosine_final_lr_ratio in last epoch</span>
    576. 

  576.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="n">lr</span> <span class="o">*</span> <span class="p">(</span><span class="mi">1</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">cosine_final_lr_ratio</span><span class="p">)</span> <span class="o">+</span> <span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">cosine_final_lr_ratio</span><span class="p">)</span>
    577. 

  577.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span></div>
    578. 

  578. 

  579. <div class="viewcode-block" id="CosineLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.CosineLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    580. 

  580.         <span class="n">post_warmup_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    581. 

  581.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">&lt;</span> <span class="n">post_warmup_epochs</span></div></div>
    582. 

  582. 

  583. 

  584. <div class="viewcode-block" id="FunctionLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.FunctionLRCallback">[docs]</a><span class="k">class</span> <span class="nc">FunctionLRCallback</span><span class="p">(</span><span class="n">LRCallbackBase</span><span class="p">):</span>
    585. 

  585.     <span class="sd">&quot;&quot;&quot;</span>
    586. 

  586. <span class="sd">    Hard coded rate scheduling for user defined lr scheduling function.</span>
    587. 

  587. <span class="sd">    &quot;&quot;&quot;</span>
    588. 

  588. 

  589.     <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">max_epochs</span><span class="p">,</span> <span class="n">lr_schedule_function</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
    590. 

  590.         <span class="nb">super</span><span class="p">(</span><span class="n">FunctionLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_BATCH_STEP</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
    591. 

  591.         <span class="k">assert</span> <span class="n">callable</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">lr_schedule_function</span><span class="p">),</span> <span class="s2">&quot;self.lr_function must be callable&quot;</span>
    592. 

  592.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_schedule_function</span> <span class="o">=</span> <span class="n">lr_schedule_function</span>
    593. 

  593.         <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">=</span> <span class="n">max_epochs</span>
    594. 

  594. 

  595. <div class="viewcode-block" id="FunctionLRCallback.is_lr_scheduling_enabled"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.FunctionLRCallback.is_lr_scheduling_enabled">[docs]</a>    <span class="k">def</span> <span class="nf">is_lr_scheduling_enabled</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    596. 

  596.         <span class="n">post_warmup_epochs</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    597. 

  597.         <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">&lt;</span> <span class="n">post_warmup_epochs</span></div>
    598. 

  598. 

  599. <div class="viewcode-block" id="FunctionLRCallback.perform_scheduling"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.FunctionLRCallback.perform_scheduling">[docs]</a>    <span class="k">def</span> <span class="nf">perform_scheduling</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">):</span>
    600. 

  600.         <span class="n">effective_epoch</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span>
    601. 

  601.         <span class="n">effective_max_epochs</span> <span class="o">=</span> <span class="p">(</span>
    602. 

  602.             <span class="bp">self</span><span class="o">.</span><span class="n">max_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">-</span> <span class="bp">self</span><span class="o">.</span><span class="n">training_params</span><span class="o">.</span><span class="n">lr_cooldown_epochs</span>
    603. 

  603.         <span class="p">)</span>
    604. 

  604.         <span class="bp">self</span><span class="o">.</span><span class="n">lr</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">lr_schedule_function</span><span class="p">(</span>
    605. 

  605.             <span class="n">initial_lr</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">initial_lr</span><span class="p">,</span>
    606. 

  606.             <span class="n">epoch</span><span class="o">=</span><span class="n">effective_epoch</span><span class="p">,</span>
    607. 

  607.             <span class="nb">iter</span><span class="o">=</span><span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">,</span>
    608. 

  608.             <span class="n">max_epoch</span><span class="o">=</span><span class="n">effective_max_epochs</span><span class="p">,</span>
    609. 

  609.             <span class="n">iters_per_epoch</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">train_loader_len</span><span class="p">,</span>
    610. 

  610.         <span class="p">)</span>
    611. 

  611.         <span class="bp">self</span><span class="o">.</span><span class="n">update_lr</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span></div></div>
    612. 

  612. 

  613. 

  614. <div class="viewcode-block" id="IllegalLRSchedulerMetric"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.IllegalLRSchedulerMetric">[docs]</a><span class="k">class</span> <span class="nc">IllegalLRSchedulerMetric</span><span class="p">(</span><span class="ne">Exception</span><span class="p">):</span>
    615. 

  615.     <span class="sd">&quot;&quot;&quot;Exception raised illegal combination of training parameters.</span>
    616. 

  616. 

  617. <span class="sd">    Attributes:</span>
    618. 

  618. <span class="sd">        message -- explanation of the error</span>
    619. 

  619. <span class="sd">    &quot;&quot;&quot;</span>
    620. 

  620. 

  621.     <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">metric_name</span><span class="p">,</span> <span class="n">metrics_dict</span><span class="p">):</span>
    622. 

  622.         <span class="bp">self</span><span class="o">.</span><span class="n">message</span> <span class="o">=</span> <span class="p">(</span>
    623. 

  623.             <span class="s2">&quot;Illegal metric name: &quot;</span> <span class="o">+</span> <span class="n">metric_name</span> <span class="o">+</span> <span class="s2">&quot;. Expected one of metics_dics keys: &quot;</span> <span class="o">+</span> <span class="nb">str</span><span class="p">(</span><span class="n">metrics_dict</span><span class="o">.</span><span class="n">keys</span><span class="p">())</span>
    624. 

  624.         <span class="p">)</span>
    625. 

  625.         <span class="nb">super</span><span class="p">()</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">message</span><span class="p">)</span></div>
    626. 

  626. 

  627. 

  628. <div class="viewcode-block" id="LRSchedulerCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.LRSchedulerCallback">[docs]</a><span class="k">class</span> <span class="nc">LRSchedulerCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    629. 

  629.     <span class="sd">&quot;&quot;&quot;</span>
    630. 

  630. <span class="sd">    Learning rate scheduler callback.</span>
    631. 

  631. 

  632. <span class="sd">    Attributes:</span>
    633. 

  633. <span class="sd">        scheduler: torch.optim._LRScheduler, the learning rate scheduler to be called step() with.</span>
    634. 

  634. <span class="sd">        metric_name: str, (default=None) the metric name for ReduceLROnPlateau learning rate scheduler.</span>
    635. 

  635. 

  636. <span class="sd">        When passing __call__ a metrics_dict, with a key=self.metric_name, the value of that metric will monitored</span>
    637. 

  637. <span class="sd">         for ReduceLROnPlateau (i.e step(metrics_dict[self.metric_name]).</span>
    638. 

  638. <span class="sd">    &quot;&quot;&quot;</span>
    639. 

  639. 

  640.     <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">scheduler</span><span class="p">,</span> <span class="n">phase</span><span class="p">,</span> <span class="n">metric_name</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
    641. 

  641.         <span class="nb">super</span><span class="p">(</span><span class="n">LRSchedulerCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    642. 

  642.         <span class="bp">self</span><span class="o">.</span><span class="n">scheduler</span> <span class="o">=</span> <span class="n">scheduler</span>
    643. 

  643.         <span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span> <span class="o">=</span> <span class="n">metric_name</span>
    644. 

  644. 

  645.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    646. 

  646.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">lr_warmup_epochs</span> <span class="o">&lt;=</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">:</span>
    647. 

  647.             <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span> <span class="ow">and</span> <span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span> <span class="ow">in</span> <span class="n">context</span><span class="o">.</span><span class="n">metrics_dict</span><span class="o">.</span><span class="n">keys</span><span class="p">():</span>
    648. 

  648.                 <span class="bp">self</span><span class="o">.</span><span class="n">scheduler</span><span class="o">.</span><span class="n">step</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">metrics_dict</span><span class="p">[</span><span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span><span class="p">])</span>
    649. 

  649.             <span class="k">elif</span> <span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
    650. 

  650.                 <span class="bp">self</span><span class="o">.</span><span class="n">scheduler</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
    651. 

  651.             <span class="k">else</span><span class="p">:</span>
    652. 

  652.                 <span class="k">raise</span> <span class="n">IllegalLRSchedulerMetric</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">metric_name</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">metrics_dict</span><span class="p">)</span>
    653. 

  653. 

  654.     <span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
    655. 

  655.         <span class="k">return</span> <span class="s2">&quot;LRSchedulerCallback: &quot;</span> <span class="o">+</span> <span class="nb">repr</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">scheduler</span><span class="p">)</span></div>
    656. 

  656. 

  657. 

  658. <div class="viewcode-block" id="MetricsUpdateCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.MetricsUpdateCallback">[docs]</a><span class="k">class</span> <span class="nc">MetricsUpdateCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    659. 

  659.     <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">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">):</span>
    660. 

  660.         <span class="nb">super</span><span class="p">(</span><span class="n">MetricsUpdateCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    661. 

  661. 

  662.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    663. 

  663.         <span class="n">context</span><span class="o">.</span><span class="n">metrics_compute_fn</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="o">**</span><span class="n">context</span><span class="o">.</span><span class="vm">__dict__</span><span class="p">)</span>
    664. 

  664.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">criterion</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
    665. 

  665.             <span class="n">context</span><span class="o">.</span><span class="n">loss_avg_meter</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">loss_log_items</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">inputs</span><span class="p">))</span></div>
    666. 

  666. 

  667. 

  668. <div class="viewcode-block" id="KDModelMetricsUpdateCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.KDModelMetricsUpdateCallback">[docs]</a><span class="k">class</span> <span class="nc">KDModelMetricsUpdateCallback</span><span class="p">(</span><span class="n">MetricsUpdateCallback</span><span class="p">):</span>
    669. 

  669.     <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">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">):</span>
    670. 

  670.         <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">phase</span><span class="o">=</span><span class="n">phase</span><span class="p">)</span>
    671. 

  671. 

  672.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    673. 

  673.         <span class="n">metrics_compute_fn_kwargs</span> <span class="o">=</span> <span class="p">{</span><span class="n">k</span><span class="p">:</span> <span class="n">v</span><span class="o">.</span><span class="n">student_output</span> <span class="k">if</span> <span class="n">k</span> <span class="o">==</span> <span class="s2">&quot;preds&quot;</span> <span class="k">else</span> <span class="n">v</span> <span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">context</span><span class="o">.</span><span class="vm">__dict__</span><span class="o">.</span><span class="n">items</span><span class="p">()}</span>
    674. 

  674.         <span class="n">context</span><span class="o">.</span><span class="n">metrics_compute_fn</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="o">**</span><span class="n">metrics_compute_fn_kwargs</span><span class="p">)</span>
    675. 

  675.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">criterion</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
    676. 

  676.             <span class="n">context</span><span class="o">.</span><span class="n">loss_avg_meter</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">loss_log_items</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">inputs</span><span class="p">))</span></div>
    677. 

  677. 

  678. 

  679. <div class="viewcode-block" id="PhaseContextTestCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.PhaseContextTestCallback">[docs]</a><span class="k">class</span> <span class="nc">PhaseContextTestCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    680. 

  680.     <span class="sd">&quot;&quot;&quot;</span>
    681. 

  681. <span class="sd">    A callback that saves the phase context the for testing.</span>
    682. 

  682. <span class="sd">    &quot;&quot;&quot;</span>
    683. 

  683. 

  684.     <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">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">):</span>
    685. 

  685.         <span class="nb">super</span><span class="p">(</span><span class="n">PhaseContextTestCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    686. 

  686.         <span class="bp">self</span><span class="o">.</span><span class="n">context</span> <span class="o">=</span> <span class="kc">None</span>
    687. 

  687. 

  688.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    689. 

  689.         <span class="bp">self</span><span class="o">.</span><span class="n">context</span> <span class="o">=</span> <span class="n">context</span></div>
    690. 

  690. 

  691. 

  692. <div class="viewcode-block" id="DetectionVisualizationCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.DetectionVisualizationCallback">[docs]</a><span class="k">class</span> <span class="nc">DetectionVisualizationCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    693. 

  693.     <span class="sd">&quot;&quot;&quot;</span>
    694. 

  694. <span class="sd">    A callback that adds a visualization of a batch of detection predictions to context.sg_logger</span>
    695. 

  695. <span class="sd">    Attributes:</span>
    696. 

  696. <span class="sd">        freq: frequency (in epochs) to perform this callback.</span>
    697. 

  697. <span class="sd">        batch_idx: batch index to perform visualization for.</span>
    698. 

  698. <span class="sd">        classes: class list of the dataset.</span>
    699. 

  699. <span class="sd">        last_img_idx_in_batch: Last image index to add to log. (default=-1, will take entire batch).</span>
    700. 

  700. <span class="sd">    &quot;&quot;&quot;</span>
    701. 

  701. 

  702.     <span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span>
    703. 

  703.         <span class="bp">self</span><span class="p">,</span>
    704. 

  704.         <span class="n">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">,</span>
    705. 

  705.         <span class="n">freq</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span>
    706. 

  706.         <span class="n">post_prediction_callback</span><span class="p">:</span> <span class="n">DetectionPostPredictionCallback</span><span class="p">,</span>
    707. 

  707.         <span class="n">classes</span><span class="p">:</span> <span class="nb">list</span><span class="p">,</span>
    708. 

  708.         <span class="n">batch_idx</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">0</span><span class="p">,</span>
    709. 

  709.         <span class="n">last_img_idx_in_batch</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span>
    710. 

  710.     <span class="p">):</span>
    711. 

  711.         <span class="nb">super</span><span class="p">(</span><span class="n">DetectionVisualizationCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    712. 

  712.         <span class="bp">self</span><span class="o">.</span><span class="n">freq</span> <span class="o">=</span> <span class="n">freq</span>
    713. 

  713.         <span class="bp">self</span><span class="o">.</span><span class="n">post_prediction_callback</span> <span class="o">=</span> <span class="n">post_prediction_callback</span>
    714. 

  714.         <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span> <span class="o">=</span> <span class="n">batch_idx</span>
    715. 

  715.         <span class="bp">self</span><span class="o">.</span><span class="n">classes</span> <span class="o">=</span> <span class="n">classes</span>
    716. 

  716.         <span class="bp">self</span><span class="o">.</span><span class="n">last_img_idx_in_batch</span> <span class="o">=</span> <span class="n">last_img_idx_in_batch</span>
    717. 

  717. 

  718.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    719. 

  719.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">%</span> <span class="bp">self</span><span class="o">.</span><span class="n">freq</span> <span class="o">==</span> <span class="mi">0</span> <span class="ow">and</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span> <span class="o">==</span> <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">:</span>
    720. 

  720.             <span class="c1"># SOME CALCULATIONS ARE IN-PLACE IN NMS, SO CLONE THE PREDICTIONS</span>
    721. 

  721.             <span class="n">preds</span> <span class="o">=</span> <span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">preds</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">clone</span><span class="p">(),</span> <span class="kc">None</span><span class="p">)</span>
    722. 

  722.             <span class="n">preds</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">post_prediction_callback</span><span class="p">(</span><span class="n">preds</span><span class="p">)</span>
    723. 

  723.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="n">DetectionVisualization</span><span class="o">.</span><span class="n">visualize_batch</span><span class="p">(</span>
    724. 

  724.                 <span class="n">context</span><span class="o">.</span><span class="n">inputs</span><span class="p">,</span> <span class="n">preds</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">target</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">classes</span>
    725. 

  725.             <span class="p">)</span>
    726. 

  726.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="p">[</span><span class="n">cv2</span><span class="o">.</span><span class="n">cvtColor</span><span class="p">(</span><span class="n">image</span><span class="p">,</span> <span class="n">cv2</span><span class="o">.</span><span class="n">COLOR_BGR2RGB</span><span class="p">)</span> <span class="k">for</span> <span class="n">image</span> <span class="ow">in</span> <span class="n">batch_imgs</span><span class="p">]</span>
    727. 

  727.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">stack</span><span class="p">(</span><span class="n">batch_imgs</span><span class="p">)</span>
    728. 

  728.             <span class="n">tag</span> <span class="o">=</span> <span class="s2">&quot;batch_&quot;</span> <span class="o">+</span> <span class="nb">str</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span> <span class="o">+</span> <span class="s2">&quot;_images&quot;</span>
    729. 

  729.             <span class="n">context</span><span class="o">.</span><span class="n">sg_logger</span><span class="o">.</span><span class="n">add_images</span><span class="p">(</span>
    730. 

  730.                 <span class="n">tag</span><span class="o">=</span><span class="n">tag</span><span class="p">,</span> <span class="n">images</span><span class="o">=</span><span class="n">batch_imgs</span><span class="p">[:</span> <span class="bp">self</span><span class="o">.</span><span class="n">last_img_idx_in_batch</span><span class="p">],</span> <span class="n">global_step</span><span class="o">=</span><span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">data_format</span><span class="o">=</span><span class="s2">&quot;NHWC&quot;</span>
    731. 

  731.             <span class="p">)</span></div>
    732. 

  732. 

  733. 

  734. <div class="viewcode-block" id="BinarySegmentationVisualizationCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.BinarySegmentationVisualizationCallback">[docs]</a><span class="k">class</span> <span class="nc">BinarySegmentationVisualizationCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    735. 

  735.     <span class="sd">&quot;&quot;&quot;</span>
    736. 

  736. <span class="sd">    A callback that adds a visualization of a batch of segmentation predictions to context.sg_logger</span>
    737. 

  737. <span class="sd">    Attributes:</span>
    738. 

  738. <span class="sd">        freq: frequency (in epochs) to perform this callback.</span>
    739. 

  739. <span class="sd">        batch_idx: batch index to perform visualization for.</span>
    740. 

  740. <span class="sd">        last_img_idx_in_batch: Last image index to add to log. (default=-1, will take entire batch).</span>
    741. 

  741. <span class="sd">    &quot;&quot;&quot;</span>
    742. 

  742. 

  743.     <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">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">,</span> <span class="n">freq</span><span class="p">:</span> <span class="nb">int</span><span class="p">,</span> <span class="n">batch_idx</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">0</span><span class="p">,</span> <span class="n">last_img_idx_in_batch</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="o">-</span><span class="mi">1</span><span class="p">):</span>
    744. 

  744.         <span class="nb">super</span><span class="p">(</span><span class="n">BinarySegmentationVisualizationCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="p">)</span>
    745. 

  745.         <span class="bp">self</span><span class="o">.</span><span class="n">freq</span> <span class="o">=</span> <span class="n">freq</span>
    746. 

  746.         <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span> <span class="o">=</span> <span class="n">batch_idx</span>
    747. 

  747.         <span class="bp">self</span><span class="o">.</span><span class="n">last_img_idx_in_batch</span> <span class="o">=</span> <span class="n">last_img_idx_in_batch</span>
    748. 

  748. 

  749.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    750. 

  750.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">%</span> <span class="bp">self</span><span class="o">.</span><span class="n">freq</span> <span class="o">==</span> <span class="mi">0</span> <span class="ow">and</span> <span class="n">context</span><span class="o">.</span><span class="n">batch_idx</span> <span class="o">==</span> <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">:</span>
    751. 

  751.             <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">preds</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">):</span>
    752. 

  752.                 <span class="n">preds</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">preds</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">clone</span><span class="p">()</span>
    753. 

  753.             <span class="k">else</span><span class="p">:</span>
    754. 

  754.                 <span class="n">preds</span> <span class="o">=</span> <span class="n">context</span><span class="o">.</span><span class="n">preds</span><span class="o">.</span><span class="n">clone</span><span class="p">()</span>
    755. 

  755.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="n">BinarySegmentationVisualization</span><span class="o">.</span><span class="n">visualize_batch</span><span class="p">(</span>
    756. 

  756.                 <span class="n">context</span><span class="o">.</span><span class="n">inputs</span><span class="p">,</span> <span class="n">preds</span><span class="p">,</span> <span class="n">context</span><span class="o">.</span><span class="n">target</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span>
    757. 

  757.             <span class="p">)</span>
    758. 

  758.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="p">[</span><span class="n">cv2</span><span class="o">.</span><span class="n">cvtColor</span><span class="p">(</span><span class="n">image</span><span class="p">,</span> <span class="n">cv2</span><span class="o">.</span><span class="n">COLOR_BGR2RGB</span><span class="p">)</span> <span class="k">for</span> <span class="n">image</span> <span class="ow">in</span> <span class="n">batch_imgs</span><span class="p">]</span>
    759. 

  759.             <span class="n">batch_imgs</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">stack</span><span class="p">(</span><span class="n">batch_imgs</span><span class="p">)</span>
    760. 

  760.             <span class="n">tag</span> <span class="o">=</span> <span class="s2">&quot;batch_&quot;</span> <span class="o">+</span> <span class="nb">str</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">batch_idx</span><span class="p">)</span> <span class="o">+</span> <span class="s2">&quot;_images&quot;</span>
    761. 

  761.             <span class="n">context</span><span class="o">.</span><span class="n">sg_logger</span><span class="o">.</span><span class="n">add_images</span><span class="p">(</span><span class="n">tag</span><span class="o">=</span><span class="n">tag</span><span class="p">,</span> <span class="n">images</span><span class="o">=</span><span class="n">batch_imgs</span><span class="p">[:</span><span class="bp">self</span><span class="o">.</span><span class="n">last_img_idx_in_batch</span><span class="p">],</span>
    762. 

  762.                                          <span class="n">global_step</span><span class="o">=</span><span class="n">context</span><span class="o">.</span><span class="n">epoch</span><span class="p">,</span> <span class="n">data_format</span><span class="o">=</span><span class="s1">&#39;NHWC&#39;</span><span class="p">)</span></div>
    763. 

  763. 

  764. 

  765. <div class="viewcode-block" id="TrainingStageSwitchCallbackBase"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.TrainingStageSwitchCallbackBase">[docs]</a><span class="k">class</span> <span class="nc">TrainingStageSwitchCallbackBase</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    766. 

  766.     <span class="sd">&quot;&quot;&quot;</span>
    767. 

  767. <span class="sd">    TrainingStageSwitchCallback</span>
    768. 

  768. 

  769. <span class="sd">    A phase callback that is called at a specific epoch (epoch start) to support multi-stage training.</span>
    770. 

  770. <span class="sd">    It does so by manipulating the objects inside the context.</span>
    771. 

  771. 

  772. <span class="sd">    Attributes:</span>
    773. 

  773. <span class="sd">        next_stage_start_epoch: int, the epoch idx to apply the stage change.</span>
    774. 

  774. <span class="sd">    &quot;&quot;&quot;</span>
    775. 

  775. 

  776.     <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">next_stage_start_epoch</span><span class="p">:</span> <span class="nb">int</span><span class="p">):</span>
    777. 

  777.         <span class="nb">super</span><span class="p">(</span><span class="n">TrainingStageSwitchCallbackBase</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">phase</span><span class="o">=</span><span class="n">Phase</span><span class="o">.</span><span class="n">TRAIN_EPOCH_START</span><span class="p">)</span>
    778. 

  778.         <span class="bp">self</span><span class="o">.</span><span class="n">next_stage_start_epoch</span> <span class="o">=</span> <span class="n">next_stage_start_epoch</span>
    779. 

  779. 

  780.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    781. 

  781.         <span class="k">if</span> <span class="n">context</span><span class="o">.</span><span class="n">epoch</span> <span class="o">==</span> <span class="bp">self</span><span class="o">.</span><span class="n">next_stage_start_epoch</span><span class="p">:</span>
    782. 

  782.             <span class="bp">self</span><span class="o">.</span><span class="n">apply_stage_change</span><span class="p">(</span><span class="n">context</span><span class="p">)</span>
    783. 

  783. 

  784. <div class="viewcode-block" id="TrainingStageSwitchCallbackBase.apply_stage_change"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.TrainingStageSwitchCallbackBase.apply_stage_change">[docs]</a>    <span class="k">def</span> <span class="nf">apply_stage_change</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    785. 

  785.         <span class="sd">&quot;&quot;&quot;</span>
    786. 

  786. <span class="sd">        This method is called when the callback is fired on the next_stage_start_epoch,</span>
    787. 

  787. <span class="sd">         and holds the stage change logic that should be applied to the context&#39;s objects.</span>
    788. 

  788. 

  789. <span class="sd">        :param context: PhaseContext, context of current phase</span>
    790. 

  790. <span class="sd">        &quot;&quot;&quot;</span>
    791. 

  791.         <span class="k">raise</span> <span class="ne">NotImplementedError</span></div></div>
    792. 

  792. 

  793. 

  794. <div class="viewcode-block" id="YoloXTrainingStageSwitchCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.YoloXTrainingStageSwitchCallback">[docs]</a><span class="k">class</span> <span class="nc">YoloXTrainingStageSwitchCallback</span><span class="p">(</span><span class="n">TrainingStageSwitchCallbackBase</span><span class="p">):</span>
    795. 

  795.     <span class="sd">&quot;&quot;&quot;</span>
    796. 

  796. <span class="sd">    YoloXTrainingStageSwitchCallback</span>
    797. 

  797. 

  798. <span class="sd">    Training stage switch for YoloX training.</span>
    799. 

  799. <span class="sd">    Disables mosaic, and manipulates YoloX loss to use L1.</span>
    800. 

  800. 

  801. <span class="sd">    &quot;&quot;&quot;</span>
    802. 

  802. 

  803.     <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">next_stage_start_epoch</span><span class="p">:</span> <span class="nb">int</span> <span class="o">=</span> <span class="mi">285</span><span class="p">):</span>
    804. 

  804.         <span class="nb">super</span><span class="p">(</span><span class="n">YoloXTrainingStageSwitchCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">next_stage_start_epoch</span><span class="o">=</span><span class="n">next_stage_start_epoch</span><span class="p">)</span>
    805. 

  805. 

  806. <div class="viewcode-block" id="YoloXTrainingStageSwitchCallback.apply_stage_change"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.YoloXTrainingStageSwitchCallback.apply_stage_change">[docs]</a>    <span class="k">def</span> <span class="nf">apply_stage_change</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    807. 

  807.         <span class="k">for</span> <span class="n">transform</span> <span class="ow">in</span> <span class="n">context</span><span class="o">.</span><span class="n">train_loader</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">transforms</span><span class="p">:</span>
    808. 

  808.             <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">transform</span><span class="p">,</span> <span class="s2">&quot;close&quot;</span><span class="p">):</span>
    809. 

  809.                 <span class="n">transform</span><span class="o">.</span><span class="n">close</span><span class="p">()</span>
    810. 

  810.         <span class="nb">iter</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">train_loader</span><span class="p">)</span>
    811. 

  811.         <span class="n">context</span><span class="o">.</span><span class="n">criterion</span><span class="o">.</span><span class="n">use_l1</span> <span class="o">=</span> <span class="kc">True</span></div></div>
    812. 

  812. 

  813. 

  814. <div class="viewcode-block" id="CallbackHandler"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.CallbackHandler">[docs]</a><span class="k">class</span> <span class="nc">CallbackHandler</span><span class="p">:</span>
    815. 

  815.     <span class="sd">&quot;&quot;&quot;</span>
    816. 

  816. <span class="sd">    Runs all callbacks who&#39;s phase attribute equals to the given phase.</span>
    817. 

  817. 

  818. <span class="sd">    Attributes:</span>
    819. 

  819. <span class="sd">        callbacks: List[PhaseCallback]. Callbacks to be run.</span>
    820. 

  820. <span class="sd">    &quot;&quot;&quot;</span>
    821. 

  821. 

  822.     <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">callbacks</span><span class="p">):</span>
    823. 

  823.         <span class="bp">self</span><span class="o">.</span><span class="n">callbacks</span> <span class="o">=</span> <span class="n">callbacks</span>
    824. 

  824. 

  825.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">phase</span><span class="p">:</span> <span class="n">Phase</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    826. 

  826.         <span class="k">for</span> <span class="n">callback</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">callbacks</span><span class="p">:</span>
    827. 

  827.             <span class="k">if</span> <span class="n">callback</span><span class="o">.</span><span class="n">phase</span> <span class="o">==</span> <span class="n">phase</span><span class="p">:</span>
    828. 

  828.                 <span class="n">callback</span><span class="p">(</span><span class="n">context</span><span class="p">)</span></div>
    829. 

  829. 

  830. 

  831. <span class="c1"># DICT FOR LEGACY LR HARD-CODED REGIMES, WILL BE DELETED IN THE FUTURE</span>
    832. 

  832. <span class="n">LR_SCHEDULERS_CLS_DICT</span> <span class="o">=</span> <span class="p">{</span>
    833. 

  833.     <span class="s2">&quot;step&quot;</span><span class="p">:</span> <span class="n">StepLRCallback</span><span class="p">,</span>
    834. 

  834.     <span class="s2">&quot;poly&quot;</span><span class="p">:</span> <span class="n">PolyLRCallback</span><span class="p">,</span>
    835. 

  835.     <span class="s2">&quot;cosine&quot;</span><span class="p">:</span> <span class="n">CosineLRCallback</span><span class="p">,</span>
    836. 

  836.     <span class="s2">&quot;exp&quot;</span><span class="p">:</span> <span class="n">ExponentialLRCallback</span><span class="p">,</span>
    837. 

  837.     <span class="s2">&quot;function&quot;</span><span class="p">:</span> <span class="n">FunctionLRCallback</span><span class="p">,</span>
    838. 

  838. <span class="p">}</span>
    839. 

  839. 

  840. <span class="n">LR_WARMUP_CLS_DICT</span> <span class="o">=</span> <span class="p">{</span><span class="s2">&quot;linear_step&quot;</span><span class="p">:</span> <span class="n">WarmupLRCallback</span><span class="p">}</span>
    841. 

  841. 

  842. 

  843. <div class="viewcode-block" id="TestLRCallback"><a class="viewcode-back" href="../../../../super_gradients.training.utils.html#super_gradients.training.utils.callbacks.TestLRCallback">[docs]</a><span class="k">class</span> <span class="nc">TestLRCallback</span><span class="p">(</span><span class="n">PhaseCallback</span><span class="p">):</span>
    844. 

  844.     <span class="sd">&quot;&quot;&quot;</span>
    845. 

  845. <span class="sd">    Phase callback that collects the learning rates in lr_placeholder at the end of each epoch (used for testing). In</span>
    846. 

  846. <span class="sd">     the case of multiple parameter groups (i.e multiple learning rates) the learning rate is collected from the first</span>
    847. 

  847. <span class="sd">     one. The phase is VALIDATION_EPOCH_END to ensure all lr updates have been performed before calling this callback.</span>
    848. 

  848. <span class="sd">    &quot;&quot;&quot;</span>
    849. 

  849. 

  850.     <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">lr_placeholder</span><span class="p">):</span>
    851. 

  851.         <span class="nb">super</span><span class="p">(</span><span class="n">TestLRCallback</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">(</span><span class="n">Phase</span><span class="o">.</span><span class="n">VALIDATION_EPOCH_END</span><span class="p">)</span>
    852. 

  852.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_placeholder</span> <span class="o">=</span> <span class="n">lr_placeholder</span>
    853. 

  853. 

  854.     <span class="k">def</span> <span class="fm">__call__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">context</span><span class="p">:</span> <span class="n">PhaseContext</span><span class="p">):</span>
    855. 

  855.         <span class="bp">self</span><span class="o">.</span><span class="n">lr_placeholder</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">context</span><span class="o">.</span><span class="n">optimizer</span><span class="o">.</span><span class="n">param_groups</span><span class="p">[</span><span class="mi">0</span><span class="p">][</span><span class="s2">&quot;lr&quot;</span><span class="p">])</span></div>
    856. 

  856. </pre></div>
    857. 

  857. 

  858.            </div>
    859. 

  859.           </div>
    860. 

  860.           <footer>
    861. 

  861. 

  862.   <hr/>
    863. 

  863. 

  864.   <div role="contentinfo">
    865. 

  865.     <p>&#169; Copyright 2021, SuperGradients team.</p>
    866. 

  866.   </div>
    867. 

  867. 

  868.   Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    869. 

  869.     <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    870. 

  870.     provided by <a href="https://readthedocs.org">Read the Docs</a>.
    871. 

  871.    
    872. 

  872. 

  873. </footer>
    874. 

  874.         </div>
    875. 

  875.       </div>
    876. 

  876.     </section>
    877. 

  877.   </div>
    878. 

  878.   <script>
    879. 

  879.       jQuery(function () {
    880. 

  880.           SphinxRtdTheme.Navigation.enable(true);
    881. 

  881.       });
    882. 

  882.   </script> 
    883. 

  883. 

  884. </body>
    885. 

  885. </html>
    886.
Discard
Prev
Next
Tip!

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