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

super_gradients.training.utils.html 277 KB
Permalink History Raw

You have to be logged in to leave a comment. Sign In
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
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
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210 
  1. <!DOCTYPE html>
    2. 

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

  3. <head>
    4. 

  4.   <meta charset="utf-8" /><meta name="generator" content="Docutils 0.17.1: http://docutils.sourceforge.net/" />
    5. 

  5. 

  6.   <meta name="viewport" content="width=device-width, initial-scale=1.0" />
    7. 

  7.   <title>super_gradients.training.utils package &mdash; SuperGradients 1.0 documentation</title>
    8. 

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

  9.       <link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
    10. 

  10.       <link rel="stylesheet" href="_static/graphviz.css" type="text/css" />
    11. 

  11.   <!--[if lt IE 9]>
    12. 

  12.     <script src="_static/js/html5shiv.min.js"></script>
    13. 

  13.   <![endif]-->
    14. 

  14.   
    15. 

  15.         <script data-url_root="./" id="documentation_options" src="_static/documentation_options.js"></script>
    16. 

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

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

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

  19.     <script src="_static/js/theme.js"></script>
    20. 

  20.     <link rel="index" title="Index" href="genindex.html" />
    21. 

  21.     <link rel="search" title="Search" href="search.html" /> 
    22. 

  22. </head>
    23. 

  23. 

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

  25.   <div class="wy-grid-for-nav">
    26. 

  26.     <nav data-toggle="wy-nav-shift" class="wy-nav-side">
    27. 

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

  28.         <div class="wy-side-nav-search" >
    29. 

  29.             <a href="index.html" class="icon icon-home"> SuperGradients
    30. 

  30.           </a>
    31. 

  31. <div role="search">
    32. 

  32.   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">
    33. 

  33.     <input type="text" name="q" placeholder="Search docs" />
    34. 

  34.     <input type="hidden" name="check_keywords" value="yes" />
    35. 

  35.     <input type="hidden" name="area" value="default" />
    36. 

  36.   </form>
    37. 

  37. </div>
    38. 

  38.         </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
    39. 

  39.               <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
    40. 

  40. <ul>
    41. 

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

  61. </ul>
    62. 

  62. 

  63.         </div>
    64. 

  64.       </div>
    65. 

  65.     </nav>
    66. 

  66. 

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

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

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

  70.       </nav>
    71. 

  71. 

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

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

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

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

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

  77.       <li>super_gradients.training.utils package</li>
    78. 

  78.       <li class="wy-breadcrumbs-aside">
    79. 

  79.             <a href="_sources/super_gradients.training.utils.rst.txt" rel="nofollow"> View page source</a>
    80. 

  80.       </li>
    81. 

  81.   </ul>
    82. 

  82.   <hr/>
    83. 

  83. </div>
    84. 

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

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

  86.              
    87. 

  87.   <section id="super-gradients-training-utils-package">
    88. 

  88. <h1>super_gradients.training.utils package<a class="headerlink" href="#super-gradients-training-utils-package" title="Permalink to this headline"></a></h1>
    89. 

  89. <section id="subpackages">
    90. 

  90. <h2>Subpackages<a class="headerlink" href="#subpackages" title="Permalink to this headline"></a></h2>
    91. 

  91. <div class="toctree-wrapper compound">
    92. 

  92. <ul>
    93. 

  93. <li class="toctree-l1"><a class="reference internal" href="super_gradients.training.utils.optimizers.html">super_gradients.training.utils.optimizers package</a><ul>
    94. 

  94. <li class="toctree-l2"><a class="reference internal" href="super_gradients.training.utils.optimizers.html#submodules">Submodules</a></li>
    95. 

  95. <li class="toctree-l2"><a class="reference internal" href="super_gradients.training.utils.optimizers.html#module-super_gradients.training.utils.optimizers.rmsprop_tf">super_gradients.training.utils.optimizers.rmsprop_tf module</a></li>
    96. 

  96. <li class="toctree-l2"><a class="reference internal" href="super_gradients.training.utils.optimizers.html#module-super_gradients.training.utils.optimizers">Module contents</a></li>
    97. 

  97. </ul>
    98. 

  98. </li>
    99. 

  99. </ul>
    100. 

  100. </div>
    101. 

  101. </section>
    102. 

  102. <section id="submodules">
    103. 

  103. <h2>Submodules<a class="headerlink" href="#submodules" title="Permalink to this headline"></a></h2>
    104. 

  104. </section>
    105. 

  105. <section id="module-super_gradients.training.utils.callbacks">
    106. 

  106. <span id="super-gradients-training-utils-callbacks-module"></span><h2>super_gradients.training.utils.callbacks module<a class="headerlink" href="#module-super_gradients.training.utils.callbacks" title="Permalink to this headline"></a></h2>
    107. 

  107. <dl class="py class">
    108. 

  108. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase">
    109. 

  109. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">Phase</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#Phase"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase" title="Permalink to this definition"></a></dt>
    110. 

  110. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">enum.Enum</span></code></p>
    111. 

  111. <p>An enumeration.</p>
    112. 

  112. <dl class="py attribute">
    113. 

  113. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.PRE_TRAINING">
    114. 

  114. <span class="sig-name descname"><span class="pre">PRE_TRAINING</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'PRE_TRAINING'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.PRE_TRAINING" title="Permalink to this definition"></a></dt>
    115. 

  115. <dd></dd></dl>
    116. 

  116. 

  117. <dl class="py attribute">
    118. 

  118. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TRAIN_BATCH_END">
    119. 

  119. <span class="sig-name descname"><span class="pre">TRAIN_BATCH_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TRAIN_BATCH_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TRAIN_BATCH_END" title="Permalink to this definition"></a></dt>
    120. 

  120. <dd></dd></dl>
    121. 

  121. 

  122. <dl class="py attribute">
    123. 

  123. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TRAIN_BATCH_STEP">
    124. 

  124. <span class="sig-name descname"><span class="pre">TRAIN_BATCH_STEP</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TRAIN_BATCH_STEP'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TRAIN_BATCH_STEP" title="Permalink to this definition"></a></dt>
    125. 

  125. <dd></dd></dl>
    126. 

  126. 

  127. <dl class="py attribute">
    128. 

  128. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TRAIN_EPOCH_START">
    129. 

  129. <span class="sig-name descname"><span class="pre">TRAIN_EPOCH_START</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TRAIN_EPOCH_START'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TRAIN_EPOCH_START" title="Permalink to this definition"></a></dt>
    130. 

  130. <dd></dd></dl>
    131. 

  131. 

  132. <dl class="py attribute">
    133. 

  133. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TRAIN_EPOCH_END">
    134. 

  134. <span class="sig-name descname"><span class="pre">TRAIN_EPOCH_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TRAIN_EPOCH_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TRAIN_EPOCH_END" title="Permalink to this definition"></a></dt>
    135. 

  135. <dd></dd></dl>
    136. 

  136. 

  137. <dl class="py attribute">
    138. 

  138. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.VALIDATION_BATCH_END">
    139. 

  139. <span class="sig-name descname"><span class="pre">VALIDATION_BATCH_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'VALIDATION_BATCH_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.VALIDATION_BATCH_END" title="Permalink to this definition"></a></dt>
    140. 

  140. <dd></dd></dl>
    141. 

  141. 

  142. <dl class="py attribute">
    143. 

  143. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.VALIDATION_EPOCH_END">
    144. 

  144. <span class="sig-name descname"><span class="pre">VALIDATION_EPOCH_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'VALIDATION_EPOCH_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.VALIDATION_EPOCH_END" title="Permalink to this definition"></a></dt>
    145. 

  145. <dd></dd></dl>
    146. 

  146. 

  147. <dl class="py attribute">
    148. 

  148. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.VALIDATION_END_BEST_EPOCH">
    149. 

  149. <span class="sig-name descname"><span class="pre">VALIDATION_END_BEST_EPOCH</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'VALIDATION_END_BEST_EPOCH'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.VALIDATION_END_BEST_EPOCH" title="Permalink to this definition"></a></dt>
    150. 

  150. <dd></dd></dl>
    151. 

  151. 

  152. <dl class="py attribute">
    153. 

  153. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TEST_BATCH_END">
    154. 

  154. <span class="sig-name descname"><span class="pre">TEST_BATCH_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TEST_BATCH_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TEST_BATCH_END" title="Permalink to this definition"></a></dt>
    155. 

  155. <dd></dd></dl>
    156. 

  156. 

  157. <dl class="py attribute">
    158. 

  158. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.TEST_END">
    159. 

  159. <span class="sig-name descname"><span class="pre">TEST_END</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'TEST_END'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.TEST_END" title="Permalink to this definition"></a></dt>
    160. 

  160. <dd></dd></dl>
    161. 

  161. 

  162. <dl class="py attribute">
    163. 

  163. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.Phase.POST_TRAINING">
    164. 

  164. <span class="sig-name descname"><span class="pre">POST_TRAINING</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'POST_TRAINING'</span></em><a class="headerlink" href="#super_gradients.training.utils.callbacks.Phase.POST_TRAINING" title="Permalink to this definition"></a></dt>
    165. 

  165. <dd></dd></dl>
    166. 

  166. 

  167. </dd></dl>
    168. 

  168. 

  169. <dl class="py class">
    170. 

  170. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.PhaseContext">
    171. 

  171. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">PhaseContext</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">epoch</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_idx</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">optimizer</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metrics_dict</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">inputs</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">preds</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metrics_compute_fn</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">loss_avg_meter</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">loss_log_items</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">criterion</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">experiment_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ckpt_dir</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">net</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr_warmup_epochs</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sg_logger</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#PhaseContext"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.PhaseContext" title="Permalink to this definition"></a></dt>
    172. 

  172. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    173. 

  173. <p>Represents the input for phase callbacks, and is constantly updated after callback calls.</p>
    174. 

  174. <dl class="py method">
    175. 

  175. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.PhaseContext.update_context">
    176. 

  176. <span class="sig-name descname"><span class="pre">update_context</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#PhaseContext.update_context"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.PhaseContext.update_context" title="Permalink to this definition"></a></dt>
    177. 

  177. <dd></dd></dl>
    178. 

  178. 

  179. </dd></dl>
    180. 

  180. 

  181. <dl class="py class">
    182. 

  182. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.PhaseCallback">
    183. 

  183. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">PhaseCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.callbacks.Phase" title="super_gradients.training.utils.callbacks.Phase"><span class="pre">super_gradients.training.utils.callbacks.Phase</span></a></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#PhaseCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="Permalink to this definition"></a></dt>
    184. 

  184. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    185. 

  185. </dd></dl>
    186. 

  186. 

  187. <dl class="py class">
    188. 

  188. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.ModelConversionCheckCallback">
    189. 

  189. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">ModelConversionCheckCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_meta_data</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#ModelConversionCheckCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.ModelConversionCheckCallback" title="Permalink to this definition"></a></dt>
    190. 

  190. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    191. 

  191. <p>Pre-training callback that verifies model conversion to onnx given specified conversion parameters.</p>
    192. 

  192. <p>The model is converted, then inference is applied with onnx runtime.</p>
    193. 

  193. <p>Use this callback wit hthe same args as DeciPlatformCallback to prevent conversion fails at the end of training.</p>
    194. 

  194. <dl class="py attribute">
    195. 

  195. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.ModelConversionCheckCallback.model_meta_data">
    196. 

  196. <span class="sig-name descname"><span class="pre">model_meta_data</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.ModelConversionCheckCallback.model_meta_data" title="Permalink to this definition"></a></dt>
    197. 

  197. <dd><p>(ModelMetadata) model’s meta-data object.</p>
    198. 

  198. </dd></dl>
    199. 

  199. 

  200. <dl class="py attribute">
    201. 

  201. <dt class="sig sig-object py">
    202. 

  202. <span class="sig-name descname"><span class="pre">The</span> <span class="pre">following</span> <span class="pre">parameters</span> <span class="pre">may</span> <span class="pre">be</span> <span class="pre">passed</span> <span class="pre">as</span> <span class="pre">kwargs</span> <span class="pre">in</span> <span class="pre">order</span> <span class="pre">to</span> <span class="pre">control</span> <span class="pre">the</span> <span class="pre">conversion</span> <span class="pre">to</span> <span class="pre">onnx</span></span></dt>
    203. 

  203. <dd></dd></dl>
    204. 

  204. 

  205. </dd></dl>
    206. 

  206. 

  207. <dl class="py class">
    208. 

  208. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback">
    209. 

  209. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">DeciLabUploadCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">email</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_meta_data</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">optimization_request_form</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">password</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ckpt_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'ckpt_best.pth'</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#DeciLabUploadCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback" title="Permalink to this definition"></a></dt>
    210. 

  210. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    211. 

  211. <p>Post-training callback for uploading and optimizing a model.</p>
    212. 

  212. <dl class="py attribute">
    213. 

  213. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback.email">
    214. 

  214. <span class="sig-name descname"><span class="pre">email</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback.email" title="Permalink to this definition"></a></dt>
    215. 

  215. <dd><p>(str) username for Deci platform.</p>
    216. 

  216. </dd></dl>
    217. 

  217. 

  218. <dl class="py attribute">
    219. 

  219. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback.model_meta_data">
    220. 

  220. <span class="sig-name descname"><span class="pre">model_meta_data</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback.model_meta_data" title="Permalink to this definition"></a></dt>
    221. 

  221. <dd><p>(ModelMetadata) model’s meta-data object.</p>
    222. 

  222. </dd></dl>
    223. 

  223. 

  224. <dl class="py attribute">
    225. 

  225. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback.optimization_request_form">
    226. 

  226. <span class="sig-name descname"><span class="pre">optimization_request_form</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback.optimization_request_form" title="Permalink to this definition"></a></dt>
    227. 

  227. <dd><p>(dict) optimization request form object.</p>
    228. 

  228. </dd></dl>
    229. 

  229. 

  230. <dl class="py attribute">
    231. 

  231. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback.password">
    232. 

  232. <span class="sig-name descname"><span class="pre">password</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback.password" title="Permalink to this definition"></a></dt>
    233. 

  233. <dd><p>(str) default=None, should only be used for testing.</p>
    234. 

  234. </dd></dl>
    235. 

  235. 

  236. <dl class="py attribute">
    237. 

  237. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DeciLabUploadCallback.ckpt_name">
    238. 

  238. <span class="sig-name descname"><span class="pre">ckpt_name</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DeciLabUploadCallback.ckpt_name" title="Permalink to this definition"></a></dt>
    239. 

  239. <dd><p>(str) default=”ckpt_best” refers to the filename of the checkpoint, inside the checkpoint directory.</p>
    240. 

  240. </dd></dl>
    241. 

  241. 

  242. <dl class="py attribute">
    243. 

  243. <dt class="sig sig-object py">
    244. 

  244. <span class="sig-name descname"><span class="pre">The</span> <span class="pre">following</span> <span class="pre">parameters</span> <span class="pre">may</span> <span class="pre">be</span> <span class="pre">passed</span> <span class="pre">as</span> <span class="pre">kwargs</span> <span class="pre">in</span> <span class="pre">order</span> <span class="pre">to</span> <span class="pre">control</span> <span class="pre">the</span> <span class="pre">conversion</span> <span class="pre">to</span> <span class="pre">onnx</span></span></dt>
    245. 

  245. <dd></dd></dl>
    246. 

  246. 

  247. </dd></dl>
    248. 

  248. 

  249. <dl class="py class">
    250. 

  250. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.LRCallbackBase">
    251. 

  251. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">LRCallbackBase</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">initial_lr</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">update_param_groups</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">train_loader_len</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">net</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">training_params</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#LRCallbackBase"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="Permalink to this definition"></a></dt>
    252. 

  252. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    253. 

  253. <p>Base class for hard coded learning rate scheduling regimes, implemented as callbacks.</p>
    254. 

  254. <dl class="py method">
    255. 

  255. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.LRCallbackBase.update_lr">
    256. 

  256. <span class="sig-name descname"><span class="pre">update_lr</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">optimizer</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">epoch</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_idx</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#LRCallbackBase.update_lr"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.LRCallbackBase.update_lr" title="Permalink to this definition"></a></dt>
    257. 

  257. <dd></dd></dl>
    258. 

  258. 

  259. </dd></dl>
    260. 

  260. 

  261. <dl class="py class">
    262. 

  262. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.WarmupLRCallback">
    263. 

  263. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">WarmupLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#WarmupLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.WarmupLRCallback" title="Permalink to this definition"></a></dt>
    264. 

  264. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    265. 

  265. <p>LR scheduling callback for linear step warmup.
    266. 

  266. LR climbs from warmup_initial_lr with even steps to initial lr. When warmup_initial_lr is None- LR climb starts from</p>
    267. 

  267. <blockquote>
    268. 

  268. <div><p>initial_lr/(1+warmup_epochs).</p>
    269. 

  269. </div></blockquote>
    270. 

  270. </dd></dl>
    271. 

  271. 

  272. <dl class="py class">
    273. 

  273. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.YoloV5WarmupLRCallback">
    274. 

  274. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">YoloV5WarmupLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#YoloV5WarmupLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.YoloV5WarmupLRCallback" title="Permalink to this definition"></a></dt>
    275. 

  275. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    276. 

  276. </dd></dl>
    277. 

  277. 

  278. <dl class="py class">
    279. 

  279. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.StepLRCallback">
    280. 

  280. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">StepLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">lr_updates</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr_decay_factor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">step_lr_update_freq</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#StepLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.StepLRCallback" title="Permalink to this definition"></a></dt>
    281. 

  281. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    282. 

  282. <p>Hard coded step learning rate scheduling (i.e at specific milestones).</p>
    283. 

  283. </dd></dl>
    284. 

  284. 

  285. <dl class="py class">
    286. 

  286. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.PolyLRCallback">
    287. 

  287. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">PolyLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_epochs</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#PolyLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.PolyLRCallback" title="Permalink to this definition"></a></dt>
    288. 

  288. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    289. 

  289. <p>Hard coded polynomial decay learning rate scheduling (i.e at specific milestones).</p>
    290. 

  290. </dd></dl>
    291. 

  291. 

  292. <dl class="py class">
    293. 

  293. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.CosineLRCallback">
    294. 

  294. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">CosineLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_epochs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cosine_final_lr_ratio</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#CosineLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.CosineLRCallback" title="Permalink to this definition"></a></dt>
    295. 

  295. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    296. 

  296. <p>Hard coded step Cosine anealing learning rate scheduling.</p>
    297. 

  297. </dd></dl>
    298. 

  298. 

  299. <dl class="py class">
    300. 

  300. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.FunctionLRCallback">
    301. 

  301. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">FunctionLRCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_epochs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr_schedule_function</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#FunctionLRCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.FunctionLRCallback" title="Permalink to this definition"></a></dt>
    302. 

  302. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.LRCallbackBase" title="super_gradients.training.utils.callbacks.LRCallbackBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.LRCallbackBase</span></code></a></p>
    303. 

  303. <p>Hard coded rate scheduling for user defined lr scheduling function.</p>
    304. 

  304. </dd></dl>
    305. 

  305. 

  306. <dl class="py exception">
    307. 

  307. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.IllegalLRSchedulerMetric">
    308. 

  308. <em class="property"><span class="pre">exception</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">IllegalLRSchedulerMetric</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">metric_name</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metrics_dict</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#IllegalLRSchedulerMetric"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.IllegalLRSchedulerMetric" title="Permalink to this definition"></a></dt>
    309. 

  309. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Exception</span></code></p>
    310. 

  310. <p>Exception raised illegal combination of training parameters.</p>
    311. 

  311. <dl class="py attribute">
    312. 

  312. <dt class="sig sig-object py">
    313. 

  313. <span class="sig-name descname"><span class="pre">message</span> <span class="pre">--</span> <span class="pre">explanation</span> <span class="pre">of</span> <span class="pre">the</span> <span class="pre">error</span></span></dt>
    314. 

  314. <dd></dd></dl>
    315. 

  315. 

  316. </dd></dl>
    317. 

  317. 

  318. <dl class="py class">
    319. 

  319. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.LRSchedulerCallback">
    320. 

  320. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">LRSchedulerCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">scheduler</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">phase</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metric_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#LRSchedulerCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.LRSchedulerCallback" title="Permalink to this definition"></a></dt>
    321. 

  321. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    322. 

  322. <p>Learning rate scheduler callback.</p>
    323. 

  323. <dl class="py attribute">
    324. 

  324. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.LRSchedulerCallback.scheduler">
    325. 

  325. <span class="sig-name descname"><span class="pre">scheduler</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.LRSchedulerCallback.scheduler" title="Permalink to this definition"></a></dt>
    326. 

  326. <dd><p>torch.optim._LRScheduler, the learning rate scheduler to be called step() with.</p>
    327. 

  327. </dd></dl>
    328. 

  328. 

  329. <dl class="py attribute">
    330. 

  330. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.LRSchedulerCallback.metric_name">
    331. 

  331. <span class="sig-name descname"><span class="pre">metric_name</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.LRSchedulerCallback.metric_name" title="Permalink to this definition"></a></dt>
    332. 

  332. <dd><p>str, (default=None) the metric name for ReduceLROnPlateau learning rate scheduler.</p>
    333. 

  333. </dd></dl>
    334. 

  334. 

  335. <dl class="py attribute">
    336. 

  336. <dt class="sig sig-object py">
    337. 

  337. <span class="sig-name descname"><span class="pre">When</span> <span class="pre">passing</span> <span class="pre">__call__</span> <span class="pre">a</span> <span class="pre">metrics_dict,</span> <span class="pre">with</span> <span class="pre">a</span> <span class="pre">key=self.metric_name,</span> <span class="pre">the</span> <span class="pre">value</span> <span class="pre">of</span> <span class="pre">that</span> <span class="pre">metric</span> <span class="pre">will</span> <span class="pre">monitored</span></span></dt>
    338. 

  338. <dd><p>for ReduceLROnPlateau (i.e step(metrics_dict[self.metric_name]).</p>
    339. 

  339. </dd></dl>
    340. 

  340. 

  341. </dd></dl>
    342. 

  342. 

  343. <dl class="py class">
    344. 

  344. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.MetricsUpdateCallback">
    345. 

  345. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">MetricsUpdateCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.callbacks.Phase" title="super_gradients.training.utils.callbacks.Phase"><span class="pre">super_gradients.training.utils.callbacks.Phase</span></a></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#MetricsUpdateCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.MetricsUpdateCallback" title="Permalink to this definition"></a></dt>
    346. 

  346. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    347. 

  347. </dd></dl>
    348. 

  348. 

  349. <dl class="py class">
    350. 

  350. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.PhaseContextTestCallback">
    351. 

  351. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">PhaseContextTestCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.callbacks.Phase" title="super_gradients.training.utils.callbacks.Phase"><span class="pre">super_gradients.training.utils.callbacks.Phase</span></a></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#PhaseContextTestCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.PhaseContextTestCallback" title="Permalink to this definition"></a></dt>
    352. 

  352. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    353. 

  353. <p>A callback that saves the phase context the for testing.</p>
    354. 

  354. </dd></dl>
    355. 

  355. 

  356. <dl class="py class">
    357. 

  357. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DetectionVisualizationCallback">
    358. 

  358. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">DetectionVisualizationCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.callbacks.Phase" title="super_gradients.training.utils.callbacks.Phase"><span class="pre">super_gradients.training.utils.callbacks.Phase</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">freq</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">post_prediction_callback</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback" title="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback"><span class="pre">super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">classes</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_idx</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">last_img_idx_in_batch</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">-</span> <span class="pre">1</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#DetectionVisualizationCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.DetectionVisualizationCallback" title="Permalink to this definition"></a></dt>
    359. 

  359. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    360. 

  360. <p>A callback that adds a visualization of a batch of detection predictions to context.sg_logger
    361. 

  361. .. attribute:: freq</p>
    362. 

  362. <blockquote>
    363. 

  363. <div><p>frequency (in epochs) to perform this callback.</p>
    364. 

  364. </div></blockquote>
    365. 

  365. <dl class="py attribute">
    366. 

  366. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DetectionVisualizationCallback.batch_idx">
    367. 

  367. <span class="sig-name descname"><span class="pre">batch_idx</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DetectionVisualizationCallback.batch_idx" title="Permalink to this definition"></a></dt>
    368. 

  368. <dd><p>batch index to perform visualization for.</p>
    369. 

  369. </dd></dl>
    370. 

  370. 

  371. <dl class="py attribute">
    372. 

  372. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DetectionVisualizationCallback.classes">
    373. 

  373. <span class="sig-name descname"><span class="pre">classes</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DetectionVisualizationCallback.classes" title="Permalink to this definition"></a></dt>
    374. 

  374. <dd><p>class list of the dataset.</p>
    375. 

  375. </dd></dl>
    376. 

  376. 

  377. <dl class="py attribute">
    378. 

  378. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.DetectionVisualizationCallback.last_img_idx_in_batch">
    379. 

  379. <span class="sig-name descname"><span class="pre">last_img_idx_in_batch</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.DetectionVisualizationCallback.last_img_idx_in_batch" title="Permalink to this definition"></a></dt>
    380. 

  380. <dd><p>Last image index to add to log. (default=-1, will take entire batch).</p>
    381. 

  381. </dd></dl>
    382. 

  382. 

  383. </dd></dl>
    384. 

  384. 

  385. <dl class="py class">
    386. 

  386. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.CallbackHandler">
    387. 

  387. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.callbacks.</span></span><span class="sig-name descname"><span class="pre">CallbackHandler</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">callbacks</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/callbacks.html#CallbackHandler"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.callbacks.CallbackHandler" title="Permalink to this definition"></a></dt>
    388. 

  388. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    389. 

  389. <p>Runs all callbacks who’s phase attribute equals to the given phase.</p>
    390. 

  390. <dl class="py attribute">
    391. 

  391. <dt class="sig sig-object py" id="super_gradients.training.utils.callbacks.CallbackHandler.callbacks">
    392. 

  392. <span class="sig-name descname"><span class="pre">callbacks</span></span><a class="headerlink" href="#super_gradients.training.utils.callbacks.CallbackHandler.callbacks" title="Permalink to this definition"></a></dt>
    393. 

  393. <dd><p>List[PhaseCallback]. Callbacks to be run.</p>
    394. 

  394. </dd></dl>
    395. 

  395. 

  396. </dd></dl>
    397. 

  397. 

  398. </section>
    399. 

  399. <section id="module-super_gradients.training.utils.checkpoint_utils">
    400. 

  400. <span id="super-gradients-training-utils-checkpoint-utils-module"></span><h2>super_gradients.training.utils.checkpoint_utils module<a class="headerlink" href="#module-super_gradients.training.utils.checkpoint_utils" title="Permalink to this headline"></a></h2>
    401. 

  401. <dl class="py function">
    402. 

  402. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.get_ckpt_local_path">
    403. 

  403. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">get_ckpt_local_path</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">source_ckpt_folder_name</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">experiment_name</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ckpt_name</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_checkpoints_location</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">external_checkpoint_path</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">overwrite_local_checkpoint</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">load_weights_only</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#get_ckpt_local_path"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.get_ckpt_local_path" title="Permalink to this definition"></a></dt>
    404. 

  404. <dd><dl class="simple">
    405. 

  405. <dt>Gets the local path to the checkpoint file, which will be:</dt><dd><ul class="simple">
    406. 

  406. <li><p>By default: YOUR_REPO_ROOT/super_gradients/checkpoints/experiment_name.</p></li>
    407. 

  407. <li><dl class="simple">
    408. 

  408. <dt>if the checkpoint file is remotely located:</dt><dd><p>when overwrite_local_checkpoint=True then it will be saved in a temporary path which will be returned,
    409. 

  409. otherwise it will be downloaded to YOUR_REPO_ROOT/super_gradients/checkpoints/experiment_name and overwrite
    410. 

  410. YOUR_REPO_ROOT/super_gradients/checkpoints/experiment_name/ckpt_name if such file exists.</p>
    411. 

  411. </dd>
    412. 

  412. </dl>
    413. 

  413. </li>
    414. 

  414. <li><p>external_checkpoint_path when external_checkpoint_path != None</p></li>
    415. 

  415. </ul>
    416. 

  416. </dd>
    417. 

  417. </dl>
    418. 

  418. <p>&#64;param source_ckpt_folder_name: The folder where the checkpoint is saved. When set to None- uses the experiment_name.
    419. 

  419. &#64;param experiment_name: experiment name attr in sg_model
    420. 

  420. &#64;param ckpt_name: checkpoint filename
    421. 

  421. &#64;param model_checkpoints_location: S3, local ot URL
    422. 

  422. &#64;param external_checkpoint_path: full path to checkpoint file (that might be located outside of super_gradients/checkpoints directory)
    423. 

  423. &#64;param overwrite_local_checkpoint: whether to overwrite the checkpoint file with the same name when downloading from S3.
    424. 

  424. &#64;param load_weights_only: whether to load the network’s state dict only.
    425. 

  425. &#64;return:</p>
    426. 

  426. </dd></dl>
    427. 

  427. 

  428. <dl class="py function">
    429. 

  429. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.adaptive_load_state_dict">
    430. 

  430. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">adaptive_load_state_dict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">net</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">state_dict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">strict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#adaptive_load_state_dict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.adaptive_load_state_dict" title="Permalink to this definition"></a></dt>
    431. 

  431. <dd><p>Adaptively loads state_dict to net, by adapting the state_dict to net’s layer names first.</p>
    432. 

  432. <p>&#64;param net: (nn.Module) to load state_dict to
    433. 

  433. &#64;param state_dict: (dict) Chekpoint state_dict
    434. 

  434. &#64;param strict: (str) key matching strictness
    435. 

  435. &#64;return:</p>
    436. 

  436. </dd></dl>
    437. 

  437. 

  438. <dl class="py function">
    439. 

  439. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.read_ckpt_state_dict">
    440. 

  440. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">read_ckpt_state_dict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ckpt_path</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'cpu'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#read_ckpt_state_dict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.read_ckpt_state_dict" title="Permalink to this definition"></a></dt>
    441. 

  441. <dd></dd></dl>
    442. 

  442. 

  443. <dl class="py function">
    444. 

  444. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.adapt_state_dict_to_fit_model_layer_names">
    445. 

  445. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">adapt_state_dict_to_fit_model_layer_names</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_state_dict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">source_ckpt</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exclude</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">[]</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#adapt_state_dict_to_fit_model_layer_names"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.adapt_state_dict_to_fit_model_layer_names" title="Permalink to this definition"></a></dt>
    446. 

  446. <dd><p>Given a model state dict and source checkpoints, the method tries to correct the keys in the model_state_dict to fit
    447. 

  447. the ckpt in order to properly load the weights into the model. If unsuccessful - returns None</p>
    448. 

  448. <blockquote>
    449. 

  449. <div><dl class="field-list simple">
    450. 

  450. <dt class="field-odd">param model_state_dict</dt>
    451. 

  451. <dd class="field-odd"><p>the model state_dict</p>
    452. 

  452. </dd>
    453. 

  453. <dt class="field-even">param source_ckpt</dt>
    454. 

  454. <dd class="field-even"><p>checkpoint dict</p>
    455. 

  455. </dd>
    456. 

  456. </dl>
    457. 

  457. <p>:exclude                  optional list for excluded layers
    458. 

  458. :return: renamed checkpoint dict (if possible)</p>
    459. 

  459. </div></blockquote>
    460. 

  460. </dd></dl>
    461. 

  461. 

  462. <dl class="py function">
    463. 

  463. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.raise_informative_runtime_error">
    464. 

  464. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">raise_informative_runtime_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">state_dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">checkpoint</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exception_msg</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#raise_informative_runtime_error"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.raise_informative_runtime_error" title="Permalink to this definition"></a></dt>
    465. 

  465. <dd><p>Given a model state dict and source checkpoints, the method calls “adapt_state_dict_to_fit_model_layer_names”
    466. 

  466. and enhances the exception_msg if loading the checkpoint_dict via the conversion method is possible</p>
    467. 

  467. </dd></dl>
    468. 

  468. 

  469. <dl class="py function">
    470. 

  470. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.load_checkpoint_to_model">
    471. 

  471. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">load_checkpoint_to_model</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ckpt_local_path</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">load_backbone</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">net</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">strict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">load_weights_only</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">load_ema_as_net</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#load_checkpoint_to_model"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.load_checkpoint_to_model" title="Permalink to this definition"></a></dt>
    472. 

  472. <dd><p>Loads the state dict in ckpt_local_path to net and returns the checkpoint’s state dict.</p>
    473. 

  473. <p>&#64;param load_ema_as_net: Will load the EMA inside the checkpoint file to the network when set
    474. 

  474. &#64;param ckpt_local_path: local path to the checkpoint file
    475. 

  475. &#64;param load_backbone: whether to load the checkpoint as a backbone
    476. 

  476. &#64;param net: network to load the checkpoint to
    477. 

  477. &#64;param strict:
    478. 

  478. &#64;param load_weights_only:
    479. 

  479. &#64;return:</p>
    480. 

  480. </dd></dl>
    481. 

  481. 

  482. <dl class="py exception">
    483. 

  483. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.MissingPretrainedWeightsException">
    484. 

  484. <em class="property"><span class="pre">exception</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">MissingPretrainedWeightsException</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">desc</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#MissingPretrainedWeightsException"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.MissingPretrainedWeightsException" title="Permalink to this definition"></a></dt>
    485. 

  485. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Exception</span></code></p>
    486. 

  486. <p>Exception raised by unsupported pretrianed model.</p>
    487. 

  487. <dl class="py attribute">
    488. 

  488. <dt class="sig sig-object py">
    489. 

  489. <span class="sig-name descname"><span class="pre">message</span> <span class="pre">--</span> <span class="pre">explanation</span> <span class="pre">of</span> <span class="pre">the</span> <span class="pre">error</span></span></dt>
    490. 

  490. <dd></dd></dl>
    491. 

  491. 

  492. </dd></dl>
    493. 

  493. 

  494. <dl class="py function">
    495. 

  495. <dt class="sig sig-object py" id="super_gradients.training.utils.checkpoint_utils.load_pretrained_weights">
    496. 

  496. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.checkpoint_utils.</span></span><span class="sig-name descname"><span class="pre">load_pretrained_weights</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">architecture</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">pretrained_weights</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#load_pretrained_weights"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.checkpoint_utils.load_pretrained_weights" title="Permalink to this definition"></a></dt>
    497. 

  497. <dd><p>Loads pretrained weights from the MODEL_URLS dictionary to model
    498. 

  498. &#64;param architecture: name of the model’s architecture
    499. 

  499. &#64;param model: model to load pretrinaed weights for
    500. 

  500. &#64;param pretrained_weights: name for the pretrianed weights (i.e imagenet)
    501. 

  501. &#64;return: None</p>
    502. 

  502. </dd></dl>
    503. 

  503. 

  504. </section>
    505. 

  505. <section id="module-super_gradients.training.utils.detection_utils">
    506. 

  506. <span id="super-gradients-training-utils-detection-utils-module"></span><h2>super_gradients.training.utils.detection_utils module<a class="headerlink" href="#module-super_gradients.training.utils.detection_utils" title="Permalink to this headline"></a></h2>
    507. 

  507. <dl class="py function">
    508. 

  508. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.base_detection_collate_fn">
    509. 

  509. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">base_detection_collate_fn</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">batch</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#base_detection_collate_fn"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.base_detection_collate_fn" title="Permalink to this definition"></a></dt>
    510. 

  510. <dd><p>Batch Processing helper function for detection training/testing.
    511. 

  511. stacks the lists of images and targets into tensors and adds the image index to each target (so the targets could
    512. 

  512. later be associated to the correct images)</p>
    513. 

  513. <blockquote>
    514. 

  514. <div><dl class="field-list simple">
    515. 

  515. <dt class="field-odd">param batch</dt>
    516. 

  516. <dd class="field-odd"><p>Input batch from the Dataset __get_item__ method</p>
    517. 

  517. </dd>
    518. 

  518. <dt class="field-even">return</dt>
    519. 

  519. <dd class="field-even"><p>batch with the transformed values</p>
    520. 

  520. </dd>
    521. 

  521. </dl>
    522. 

  522. </div></blockquote>
    523. 

  523. </dd></dl>
    524. 

  524. 

  525. <dl class="py function">
    526. 

  526. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.convert_xyxy_bbox_to_xywh">
    527. 

  527. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">convert_xyxy_bbox_to_xywh</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input_bbox</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#convert_xyxy_bbox_to_xywh"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.convert_xyxy_bbox_to_xywh" title="Permalink to this definition"></a></dt>
    528. 

  528. <dd><dl class="simple">
    529. 

  529. <dt>convert_xyxy_bbox_to_xywh - Converts bounding box format from [x1, y1, x2, y2] to [x, y, w, h]</dt><dd><dl class="field-list simple">
    530. 

  530. <dt class="field-odd">param input_bbox</dt>
    531. 

  531. <dd class="field-odd"><p>input bbox</p>
    532. 

  532. </dd>
    533. 

  533. <dt class="field-even">return</dt>
    534. 

  534. <dd class="field-even"><p>Converted bbox</p>
    535. 

  535. </dd>
    536. 

  536. </dl>
    537. 

  537. </dd>
    538. 

  538. </dl>
    539. 

  539. </dd></dl>
    540. 

  540. 

  541. <dl class="py function">
    542. 

  542. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.convert_xywh_bbox_to_xyxy">
    543. 

  543. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">convert_xywh_bbox_to_xyxy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input_bbox</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#convert_xywh_bbox_to_xyxy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.convert_xywh_bbox_to_xyxy" title="Permalink to this definition"></a></dt>
    544. 

  544. <dd><dl class="simple">
    545. 

  545. <dt>Converts bounding box format from [x, y, w, h] to [x1, y1, x2, y2]</dt><dd><dl class="field-list simple">
    546. 

  546. <dt class="field-odd">param input_bbox</dt>
    547. 

  547. <dd class="field-odd"><p>input bbox either 2-dimensional (for all boxes of a single image) or 3-dimensional (for
    548. 

  548. boxes of a batch of images)</p>
    549. 

  549. </dd>
    550. 

  550. <dt class="field-even">return</dt>
    551. 

  551. <dd class="field-even"><p>Converted bbox in same dimensions as the original</p>
    552. 

  552. </dd>
    553. 

  553. </dl>
    554. 

  554. </dd>
    555. 

  555. </dl>
    556. 

  556. </dd></dl>
    557. 

  557. 

  558. <dl class="py function">
    559. 

  559. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.calculate_wh_iou">
    560. 

  560. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">calculate_wh_iou</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">box1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">box2</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">float</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#calculate_wh_iou"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.calculate_wh_iou" title="Permalink to this definition"></a></dt>
    561. 

  561. <dd><dl class="simple">
    562. 

  562. <dt>calculate_wh_iou - Gets the Intersection over Union of the w,h values of the bboxes</dt><dd><dl class="field-list simple">
    563. 

  563. <dt class="field-odd">param box1</dt>
    564. 

  564. <dd class="field-odd"><p></p></dd>
    565. 

  565. <dt class="field-even">param box2</dt>
    566. 

  566. <dd class="field-even"><p></p></dd>
    567. 

  567. <dt class="field-odd">return</dt>
    568. 

  568. <dd class="field-odd"><p>IOU</p>
    569. 

  569. </dd>
    570. 

  570. </dl>
    571. 

  571. </dd>
    572. 

  572. </dl>
    573. 

  573. </dd></dl>
    574. 

  574. 

  575. <dl class="py function">
    576. 

  576. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.calculate_bbox_iou_matrix">
    577. 

  577. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">calculate_bbox_iou_matrix</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">box1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">box2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">x1y1x2y2</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">GIoU</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">DIoU</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">CIoU</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1e-09</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#calculate_bbox_iou_matrix"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.calculate_bbox_iou_matrix" title="Permalink to this definition"></a></dt>
    578. 

  578. <dd><dl class="simple">
    579. 

  579. <dt>calculate iou matrix containing the iou of every couple iuo(i,j) where i is in box1 and j is in box2</dt><dd><dl class="field-list simple">
    580. 

  580. <dt class="field-odd">param box1</dt>
    581. 

  581. <dd class="field-odd"><p>a 2D tensor of boxes (shape N x 4)</p>
    582. 

  582. </dd>
    583. 

  583. <dt class="field-even">param box2</dt>
    584. 

  584. <dd class="field-even"><p>a 2D tensor of boxes (shape M x 4)</p>
    585. 

  585. </dd>
    586. 

  586. <dt class="field-odd">param x1y1x2y2</dt>
    587. 

  587. <dd class="field-odd"><p>boxes format is x1y1x2y2 (True) or xywh where xy is the center (False)</p>
    588. 

  588. </dd>
    589. 

  589. <dt class="field-even">return</dt>
    590. 

  590. <dd class="field-even"><p>a 2D iou matrix (shape NxM)</p>
    591. 

  591. </dd>
    592. 

  592. </dl>
    593. 

  593. </dd>
    594. 

  594. </dl>
    595. 

  595. </dd></dl>
    596. 

  596. 

  597. <dl class="py function">
    598. 

  598. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.calculate_bbox_iou_elementwise">
    599. 

  599. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">calculate_bbox_iou_elementwise</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">box1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">box2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">x1y1x2y2</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">GIoU</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">DIoU</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">CIoU</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1e-09</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#calculate_bbox_iou_elementwise"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.calculate_bbox_iou_elementwise" title="Permalink to this definition"></a></dt>
    600. 

  600. <dd><dl class="simple">
    601. 

  601. <dt>calculate elementwise iou of two bbox tensors</dt><dd><dl class="field-list simple">
    602. 

  602. <dt class="field-odd">param box1</dt>
    603. 

  603. <dd class="field-odd"><p>a 2D tensor of boxes (shape N x 4)</p>
    604. 

  604. </dd>
    605. 

  605. <dt class="field-even">param box2</dt>
    606. 

  606. <dd class="field-even"><p>a 2D tensor of boxes (shape N x 4)</p>
    607. 

  607. </dd>
    608. 

  608. <dt class="field-odd">param x1y1x2y2</dt>
    609. 

  609. <dd class="field-odd"><p>boxes format is x1y1x2y2 (True) or xywh where xy is the center (False)</p>
    610. 

  610. </dd>
    611. 

  611. <dt class="field-even">return</dt>
    612. 

  612. <dd class="field-even"><p>a 1D iou tensor (shape N)</p>
    613. 

  613. </dd>
    614. 

  614. </dl>
    615. 

  615. </dd>
    616. 

  616. </dl>
    617. 

  617. </dd></dl>
    618. 

  618. 

  619. <dl class="py function">
    620. 

  620. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.calc_bbox_iou_matrix">
    621. 

  621. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">calc_bbox_iou_matrix</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">pred</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#calc_bbox_iou_matrix"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.calc_bbox_iou_matrix" title="Permalink to this definition"></a></dt>
    622. 

  622. <dd><p>calculate iou for every pair of boxes in the boxes vector
    623. 

  623. :param pred: a 3-dimensional tensor containing all boxes for a batch of images [N, num_boxes, 4], where</p>
    624. 

  624. <blockquote>
    625. 

  625. <div><p>each box format is [x1,y1,x2,y2]</p>
    626. 

  626. </div></blockquote>
    627. 

  627. <dl class="field-list simple">
    628. 

  628. <dt class="field-odd">Returns</dt>
    629. 

  629. <dd class="field-odd"><p>a 3-dimensional matrix where M_i_j_k is the iou of box j and box k of the i’th image in the batch</p>
    630. 

  630. </dd>
    631. 

  631. </dl>
    632. 

  632. </dd></dl>
    633. 

  633. 

  634. <dl class="py function">
    635. 

  635. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.build_detection_targets">
    636. 

  636. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">build_detection_targets</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">detection_net</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#build_detection_targets"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.build_detection_targets" title="Permalink to this definition"></a></dt>
    637. 

  637. <dd><dl>
    638. 

  638. <dt>build_detection_targets - Builds the outputs of the Detection NN</dt><dd><blockquote>
    639. 

  639. <div><p>This function filters all of the targets that don’t have a sufficient iou coverage
    640. 

  640. of the Model’s pre-trained k-means anchors
    641. 

  641. The iou_threshold is a parameter of the NN Model</p>
    642. 

  642. </div></blockquote>
    643. 

  643. <dl class="field-list simple">
    644. 

  644. <dt class="field-odd">param detection_net</dt>
    645. 

  645. <dd class="field-odd"><p>The nn.Module of the Detection Algorithm</p>
    646. 

  646. </dd>
    647. 

  647. <dt class="field-even">param targets</dt>
    648. 

  648. <dd class="field-even"><p>targets (labels)</p>
    649. 

  649. </dd>
    650. 

  650. <dt class="field-odd">return</dt>
    651. 

  651. <dd class="field-odd"><p></p></dd>
    652. 

  652. </dl>
    653. 

  653. </dd>
    654. 

  654. </dl>
    655. 

  655. </dd></dl>
    656. 

  656. 

  657. <dl class="py function">
    658. 

  658. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.yolo_v3_non_max_suppression">
    659. 

  659. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">yolo_v3_non_max_suppression</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prediction</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">conf_thres</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">nms_thres</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'cpu'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#yolo_v3_non_max_suppression"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.yolo_v3_non_max_suppression" title="Permalink to this definition"></a></dt>
    660. 

  660. <dd><dl>
    661. 

  661. <dt>non_max_suppression - Removes detections with lower object confidence score than ‘conf_thres’</dt><dd><blockquote>
    662. 

  662. <div><p>Non-Maximum Suppression to further filter detections.</p>
    663. 

  663. </div></blockquote>
    664. 

  664. <dl class="field-list simple">
    665. 

  665. <dt class="field-odd">param prediction</dt>
    666. 

  666. <dd class="field-odd"><p>the raw prediction as produced by the yolo_v3 network</p>
    667. 

  667. </dd>
    668. 

  668. <dt class="field-even">param conf_thres</dt>
    669. 

  669. <dd class="field-even"><p>confidence threshold - only prediction with confidence score higher than the threshold
    670. 

  670. will be considered</p>
    671. 

  671. </dd>
    672. 

  672. <dt class="field-odd">param nms_thres</dt>
    673. 

  673. <dd class="field-odd"><p>IoU threshold for the nms algorithm</p>
    674. 

  674. </dd>
    675. 

  675. <dt class="field-even">param device</dt>
    676. 

  676. <dd class="field-even"><p>the device to move all output tensors into</p>
    677. 

  677. </dd>
    678. 

  678. <dt class="field-odd">return</dt>
    679. 

  679. <dd class="field-odd"><p>(x1, y1, x2, y2, object_conf, class_conf, class)</p>
    680. 

  680. </dd>
    681. 

  681. </dl>
    682. 

  682. </dd>
    683. 

  683. </dl>
    684. 

  684. </dd></dl>
    685. 

  685. 

  686. <dl class="py function">
    687. 

  687. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.change_bbox_bounds_for_image_size">
    688. 

  688. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">change_bbox_bounds_for_image_size</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">boxes</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">img_shape</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#change_bbox_bounds_for_image_size"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.change_bbox_bounds_for_image_size" title="Permalink to this definition"></a></dt>
    689. 

  689. <dd></dd></dl>
    690. 

  690. 

  691. <dl class="py function">
    692. 

  692. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.rescale_bboxes_for_image_size">
    693. 

  693. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">rescale_bboxes_for_image_size</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">current_image_shape</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">bbox</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">original_image_shape</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ratio_pad</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#rescale_bboxes_for_image_size"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.rescale_bboxes_for_image_size" title="Permalink to this definition"></a></dt>
    694. 

  694. <dd><dl class="simple">
    695. 

  695. <dt>rescale_bboxes_for_image_size - Changes the bboxes to fit the original image</dt><dd><dl class="field-list simple">
    696. 

  696. <dt class="field-odd">param current_image_shape</dt>
    697. 

  697. <dd class="field-odd"><p></p></dd>
    698. 

  698. <dt class="field-even">param bbox</dt>
    699. 

  699. <dd class="field-even"><p></p></dd>
    700. 

  700. <dt class="field-odd">param original_image_shape</dt>
    701. 

  701. <dd class="field-odd"><p></p></dd>
    702. 

  702. <dt class="field-even">param ratio_pad</dt>
    703. 

  703. <dd class="field-even"><p></p></dd>
    704. 

  704. <dt class="field-odd">return</dt>
    705. 

  705. <dd class="field-odd"><p></p></dd>
    706. 

  706. </dl>
    707. 

  707. </dd>
    708. 

  708. </dl>
    709. 

  709. </dd></dl>
    710. 

  710. 

  711. <dl class="py class">
    712. 

  712. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback">
    713. 

  713. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">DetectionPostPredictionCallback</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#DetectionPostPredictionCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback" title="Permalink to this definition"></a></dt>
    714. 

  714. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">abc.ABC</span></code>, <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    715. 

  715. <dl class="py method">
    716. 

  716. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback.forward">
    717. 

  717. <em class="property"><span class="pre">abstract</span> </em><span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#DetectionPostPredictionCallback.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback.forward" title="Permalink to this definition"></a></dt>
    718. 

  718. <dd><dl class="field-list simple">
    719. 

  719. <dt class="field-odd">Parameters</dt>
    720. 

  720. <dd class="field-odd"><ul class="simple">
    721. 

  721. <li><p><strong>x</strong> – the output of your model</p></li>
    722. 

  722. <li><p><strong>device</strong> – the device to move all output tensors into</p></li>
    723. 

  723. </ul>
    724. 

  724. </dd>
    725. 

  725. <dt class="field-even">Returns</dt>
    726. 

  726. <dd class="field-even"><p>a list with length batch_size, each item in the list is a detections
    727. 

  727. with shape: nx6 (x1, y1, x2, y2, confidence, class) where x and y are in range [0,1]</p>
    728. 

  728. </dd>
    729. 

  729. </dl>
    730. 

  730. </dd></dl>
    731. 

  731. 

  732. <dl class="py attribute">
    733. 

  733. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback.training">
    734. 

  734. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback.training" title="Permalink to this definition"></a></dt>
    735. 

  735. <dd></dd></dl>
    736. 

  736. 

  737. </dd></dl>
    738. 

  738. 

  739. <dl class="py class">
    740. 

  740. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression">
    741. 

  741. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">YoloV3NonMaxSuppression</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">conf</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.001</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">nms_thres</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_predictions</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">500</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#YoloV3NonMaxSuppression"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression" title="Permalink to this definition"></a></dt>
    742. 

  742. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback" title="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback</span></code></a></p>
    743. 

  743. <dl class="py method">
    744. 

  744. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression.forward">
    745. 

  745. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#YoloV3NonMaxSuppression.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression.forward" title="Permalink to this definition"></a></dt>
    746. 

  746. <dd><dl class="field-list simple">
    747. 

  747. <dt class="field-odd">Parameters</dt>
    748. 

  748. <dd class="field-odd"><ul class="simple">
    749. 

  749. <li><p><strong>x</strong> – the output of your model</p></li>
    750. 

  750. <li><p><strong>device</strong> – the device to move all output tensors into</p></li>
    751. 

  751. </ul>
    752. 

  752. </dd>
    753. 

  753. <dt class="field-even">Returns</dt>
    754. 

  754. <dd class="field-even"><p>a list with length batch_size, each item in the list is a detections
    755. 

  755. with shape: nx6 (x1, y1, x2, y2, confidence, class) where x and y are in range [0,1]</p>
    756. 

  756. </dd>
    757. 

  757. </dl>
    758. 

  758. </dd></dl>
    759. 

  759. 

  760. <dl class="py attribute">
    761. 

  761. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression.training">
    762. 

  762. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.YoloV3NonMaxSuppression.training" title="Permalink to this definition"></a></dt>
    763. 

  763. <dd></dd></dl>
    764. 

  764. 

  765. </dd></dl>
    766. 

  766. 

  767. <dl class="py class">
    768. 

  768. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.IouThreshold">
    769. 

  769. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">IouThreshold</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#IouThreshold"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.IouThreshold" title="Permalink to this definition"></a></dt>
    770. 

  770. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">tuple</span></code>, <code class="xref py py-class docutils literal notranslate"><span class="pre">enum.Enum</span></code></p>
    771. 

  771. <p>An enumeration.</p>
    772. 

  772. <dl class="py attribute">
    773. 

  773. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.IouThreshold.MAP_05">
    774. 

  774. <span class="sig-name descname"><span class="pre">MAP_05</span></span><em class="property"> <span class="pre">=</span> <span class="pre">(0.5,</span> <span class="pre">0.5)</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.IouThreshold.MAP_05" title="Permalink to this definition"></a></dt>
    775. 

  775. <dd></dd></dl>
    776. 

  776. 

  777. <dl class="py attribute">
    778. 

  778. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.IouThreshold.MAP_05_TO_095">
    779. 

  779. <span class="sig-name descname"><span class="pre">MAP_05_TO_095</span></span><em class="property"> <span class="pre">=</span> <span class="pre">(0.5,</span> <span class="pre">0.95)</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.IouThreshold.MAP_05_TO_095" title="Permalink to this definition"></a></dt>
    780. 

  780. <dd></dd></dl>
    781. 

  781. 

  782. <dl class="py method">
    783. 

  783. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.IouThreshold.is_range">
    784. 

  784. <span class="sig-name descname"><span class="pre">is_range</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#IouThreshold.is_range"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.IouThreshold.is_range" title="Permalink to this definition"></a></dt>
    785. 

  785. <dd></dd></dl>
    786. 

  786. 

  787. </dd></dl>
    788. 

  788. 

  789. <dl class="py function">
    790. 

  790. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.scale_img">
    791. 

  791. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">scale_img</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">img</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ratio</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">pad_to_original_img_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#scale_img"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.scale_img" title="Permalink to this definition"></a></dt>
    792. 

  792. <dd><p>Scales the image by ratio (image dims is (batch_size, channels, height, width)
    793. 

  793. Taken from Yolov5 Ultralitics repo</p>
    794. 

  794. </dd></dl>
    795. 

  795. 

  796. <dl class="py function">
    797. 

  797. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.fuse_conv_and_bn">
    798. 

  798. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">fuse_conv_and_bn</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">conv</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">bn</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#fuse_conv_and_bn"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.fuse_conv_and_bn" title="Permalink to this definition"></a></dt>
    799. 

  799. <dd><p>Fuse convolution and batchnorm layers <a class="reference external" href="https://tehnokv.com/posts/fusing-batchnorm-and-conv/">https://tehnokv.com/posts/fusing-batchnorm-and-conv/</a>
    800. 

  800. Taken from Yolov5 Ultralitics repo</p>
    801. 

  801. </dd></dl>
    802. 

  802. 

  803. <dl class="py function">
    804. 

  804. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.check_anchor_order">
    805. 

  805. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">check_anchor_order</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">m</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#check_anchor_order"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.check_anchor_order" title="Permalink to this definition"></a></dt>
    806. 

  806. <dd><p>Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary
    807. 

  807. Taken from Yolov5 Ultralitics repo</p>
    808. 

  808. </dd></dl>
    809. 

  809. 

  810. <dl class="py function">
    811. 

  811. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.box_iou">
    812. 

  812. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">box_iou</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">box1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">box2</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#box_iou"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.box_iou" title="Permalink to this definition"></a></dt>
    813. 

  813. <dd><p>Return intersection-over-union (Jaccard index) of boxes.
    814. 

  814. Both sets of boxes are expected to be in (x1, y1, x2, y2) format.
    815. 

  815. :param box1:
    816. 

  816. :type box1: Tensor[N, 4]
    817. 

  817. :param box2:
    818. 

  818. :type box2: Tensor[M, 4]</p>
    819. 

  819. <dl class="field-list simple">
    820. 

  820. <dt class="field-odd">Returns</dt>
    821. 

  821. <dd class="field-odd"><p><dl class="simple">
    822. 

  822. <dt>the NxM matrix containing the pairwise</dt><dd><p>IoU values for every element in boxes1 and boxes2</p>
    823. 

  823. </dd>
    824. 

  824. </dl>
    825. 

  825. </p>
    826. 

  826. </dd>
    827. 

  827. <dt class="field-even">Return type</dt>
    828. 

  828. <dd class="field-even"><p>iou (Tensor[N, M])</p>
    829. 

  829. </dd>
    830. 

  830. </dl>
    831. 

  831. <p>Taken from Yolov5 Ultralitics repo</p>
    832. 

  832. </dd></dl>
    833. 

  833. 

  834. <dl class="py function">
    835. 

  835. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.non_max_suppression">
    836. 

  836. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">non_max_suppression</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prediction</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">conf_thres</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">iou_thres</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.6</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">merge</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">classes</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">agnostic</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#non_max_suppression"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.non_max_suppression" title="Permalink to this definition"></a></dt>
    837. 

  837. <dd><dl class="simple">
    838. 

  838. <dt>Performs Non-Maximum Suppression (NMS) on inference results</dt><dd><dl class="field-list simple">
    839. 

  839. <dt class="field-odd">param prediction</dt>
    840. 

  840. <dd class="field-odd"><p>raw model prediction</p>
    841. 

  841. </dd>
    842. 

  842. <dt class="field-even">param conf_thres</dt>
    843. 

  843. <dd class="field-even"><p>below the confidence threshold - prediction are discarded</p>
    844. 

  844. </dd>
    845. 

  845. <dt class="field-odd">param iou_thres</dt>
    846. 

  846. <dd class="field-odd"><p>IoU threshold for the nms algorithm</p>
    847. 

  847. </dd>
    848. 

  848. <dt class="field-even">param merge</dt>
    849. 

  849. <dd class="field-even"><p>Merge boxes using weighted mean</p>
    850. 

  850. </dd>
    851. 

  851. <dt class="field-odd">param classes</dt>
    852. 

  852. <dd class="field-odd"><p>(optional list) filter by class</p>
    853. 

  853. </dd>
    854. 

  854. <dt class="field-even">param agnostic</dt>
    855. 

  855. <dd class="field-even"><p>Determines if is class agnostic. i.e. may display a box with 2 predictions</p>
    856. 

  856. </dd>
    857. 

  857. <dt class="field-odd">return</dt>
    858. 

  858. <dd class="field-odd"><p>(x1, y1, x2, y2, object_conf, class_conf, class)</p>
    859. 

  859. </dd>
    860. 

  860. </dl>
    861. 

  861. </dd>
    862. 

  862. </dl>
    863. 

  863. <dl class="field-list simple">
    864. 

  864. <dt class="field-odd">Returns</dt>
    865. 

  865. <dd class="field-odd"><p>nx6 (x1, y1, x2, y2, conf, cls)</p>
    866. 

  866. </dd>
    867. 

  867. <dt class="field-even">Return type</dt>
    868. 

  868. <dd class="field-even"><p>detections with shape</p>
    869. 

  869. </dd>
    870. 

  870. </dl>
    871. 

  871. </dd></dl>
    872. 

  872. 

  873. <dl class="py function">
    874. 

  874. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.check_img_size_divisibilty">
    875. 

  875. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">check_img_size_divisibilty</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">img_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stride</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">32</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#check_img_size_divisibilty"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.check_img_size_divisibilty" title="Permalink to this definition"></a></dt>
    876. 

  876. <dd><dl class="field-list simple">
    877. 

  877. <dt class="field-odd">Parameters</dt>
    878. 

  878. <dd class="field-odd"><ul class="simple">
    879. 

  879. <li><p><strong>img_size</strong> – Int, the size of the image (H or W).</p></li>
    880. 

  880. <li><p><strong>stride</strong> – Int, the number to check if img_size is divisible by.</p></li>
    881. 

  881. </ul>
    882. 

  882. </dd>
    883. 

  883. <dt class="field-even">Returns</dt>
    884. 

  884. <dd class="field-even"><p>(True, None) if img_size is divisble by stride, (False, Suggestions) if it’s not.
    885. 

  885. Note: Suggestions are the two closest numbers to img_size that <em>are</em> divisible by stride.
    886. 

  886. For example if img_size=321, stride=32, it will return (False,(352, 320)).</p>
    887. 

  887. </dd>
    888. 

  888. </dl>
    889. 

  889. </dd></dl>
    890. 

  890. 

  891. <dl class="py function">
    892. 

  892. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.make_divisible">
    893. 

  893. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">make_divisible</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">divisor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ceil</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#make_divisible"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.make_divisible" title="Permalink to this definition"></a></dt>
    894. 

  894. <dd><p>Returns x evenly divisible by divisor.
    895. 

  895. If ceil=True it will return the closest larger number to the original x, and ceil=False the closest smaller number.</p>
    896. 

  896. </dd></dl>
    897. 

  897. 

  898. <dl class="py function">
    899. 

  899. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.matrix_non_max_suppression">
    900. 

  900. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">matrix_non_max_suppression</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">pred</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">conf_thres</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">kernel</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'gaussian'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sigma</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">3.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_num_of_detections</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">500</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#matrix_non_max_suppression"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.matrix_non_max_suppression" title="Permalink to this definition"></a></dt>
    901. 

  901. <dd><dl class="simple">
    902. 

  902. <dt>Performs Matrix Non-Maximum Suppression (NMS) on inference results</dt><dd><p><a class="reference external" href="https://arxiv.org/pdf/1912.04488.pdf">https://arxiv.org/pdf/1912.04488.pdf</a>
    903. 

  903. :param pred: raw model prediction (in test mode) - a Tensor of shape [batch, num_predictions, 85]
    904. 

  904. where each item format is (x, y, w, h, object_conf, class_conf, … 80 classes score …)
    905. 

  905. :param conf_thres: below the confidence threshold - prediction are discarded
    906. 

  906. :param kernel: type of kernel to use [‘gaussian’, ‘linear’]
    907. 

  907. :param sigma: sigma for the gussian kernel
    908. 

  908. :param max_num_of_detections: maximum number of boxes to output
    909. 

  909. :return:  list of (x1, y1, x2, y2, object_conf, class_conf, class)</p>
    910. 

  910. </dd>
    911. 

  911. </dl>
    912. 

  912. <dl class="field-list simple">
    913. 

  913. <dt class="field-odd">Returns</dt>
    914. 

  914. <dd class="field-odd"><p>(x1, y1, x2, y2, conf, cls)</p>
    915. 

  915. </dd>
    916. 

  916. <dt class="field-even">Return type</dt>
    917. 

  917. <dd class="field-even"><p>detections list with shape</p>
    918. 

  918. </dd>
    919. 

  919. </dl>
    920. 

  920. </dd></dl>
    921. 

  921. 

  922. <dl class="py class">
    923. 

  923. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.NMS_Type">
    924. 

  924. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">NMS_Type</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#NMS_Type"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.NMS_Type" title="Permalink to this definition"></a></dt>
    925. 

  925. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">str</span></code>, <code class="xref py py-class docutils literal notranslate"><span class="pre">enum.Enum</span></code></p>
    926. 

  926. <p>Type of non max suppression algorithm that can be used for post processing detection</p>
    927. 

  927. <dl class="py attribute">
    928. 

  928. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.NMS_Type.ITERATIVE">
    929. 

  929. <span class="sig-name descname"><span class="pre">ITERATIVE</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'iterative'</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.NMS_Type.ITERATIVE" title="Permalink to this definition"></a></dt>
    930. 

  930. <dd></dd></dl>
    931. 

  931. 

  932. <dl class="py attribute">
    933. 

  933. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.NMS_Type.MATRIX">
    934. 

  934. <span class="sig-name descname"><span class="pre">MATRIX</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'matrix'</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.NMS_Type.MATRIX" title="Permalink to this definition"></a></dt>
    935. 

  935. <dd></dd></dl>
    936. 

  936. 

  937. </dd></dl>
    938. 

  938. 

  939. <dl class="py function">
    940. 

  940. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.calc_batch_prediction_accuracy">
    941. 

  941. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">calc_batch_prediction_accuracy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">output</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">height</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">width</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">iou_thres</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.detection_utils.IouThreshold" title="super_gradients.training.utils.detection_utils.IouThreshold"><span class="pre">super_gradients.training.utils.detection_utils.IouThreshold</span></a></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">tuple</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#calc_batch_prediction_accuracy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.calc_batch_prediction_accuracy" title="Permalink to this definition"></a></dt>
    942. 

  942. <dd><dl class="field-list simple">
    943. 

  943. <dt class="field-odd">Parameters</dt>
    944. 

  944. <dd class="field-odd"><ul class="simple">
    945. 

  945. <li><p><strong>output</strong> – list (of length batch_size) of Tensors of shape (num_detections, 6)
    946. 

  946. format:     (x1, y1, x2, y2, confidence, class_label) where x1,y1,x2,y2 are according to image size</p></li>
    947. 

  947. <li><p><strong>targets</strong> – targets for all images of shape (total_num_targets, 6)
    948. 

  948. format:     (image_index, x, y, w, h, label) where x,y,w,h are in range [0,1]</p></li>
    949. 

  949. <li><p><strong>height</strong><strong>,</strong><strong>width</strong> – dimensions of the image</p></li>
    950. 

  950. <li><p><strong>iou_thres</strong> – Threshold to compute the mAP</p></li>
    951. 

  951. <li><p><strong>device</strong> – ‘cuda’’cpu’ - where the computations are made</p></li>
    952. 

  952. </ul>
    953. 

  953. </dd>
    954. 

  954. <dt class="field-even">Returns</dt>
    955. 

  955. <dd class="field-even"><p></p>
    956. 

  956. </dd>
    957. 

  957. </dl>
    958. 

  958. </dd></dl>
    959. 

  959. 

  960. <dl class="py class">
    961. 

  961. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.AnchorGenerator">
    962. 

  962. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">AnchorGenerator</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#AnchorGenerator"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.AnchorGenerator" title="Permalink to this definition"></a></dt>
    963. 

  963. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    964. 

  964. <dl class="py attribute">
    965. 

  965. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.AnchorGenerator.logger">
    966. 

  966. <span class="sig-name descname"><span class="pre">logger</span></span><em class="property"> <span class="pre">=</span> <span class="pre">&lt;Logger</span> <span class="pre">super_gradients.training.utils.detection_utils</span> <span class="pre">(INFO)&gt;</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.AnchorGenerator.logger" title="Permalink to this definition"></a></dt>
    967. 

  967. <dd></dd></dl>
    968. 

  968. 

  969. </dd></dl>
    970. 

  970. 

  971. <dl class="py function">
    972. 

  972. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.plot_coco_datasaet_images_with_detections">
    973. 

  973. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">plot_coco_datasaet_images_with_detections</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data_loader</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_images_to_plot</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#plot_coco_datasaet_images_with_detections"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.plot_coco_datasaet_images_with_detections" title="Permalink to this definition"></a></dt>
    974. 

  974. <dd><dl class="simple">
    975. 

  975. <dt>plot_coco_images</dt><dd><dl class="field-list simple">
    976. 

  976. <dt class="field-odd">param data_loader</dt>
    977. 

  977. <dd class="field-odd"><p></p></dd>
    978. 

  978. <dt class="field-even">param num_images_to_plot</dt>
    979. 

  979. <dd class="field-even"><p></p></dd>
    980. 

  980. <dt class="field-odd">return</dt>
    981. 

  981. <dd class="field-odd"><p></p></dd>
    982. 

  982. </dl>
    983. 

  983. </dd>
    984. 

  984. </dl>
    985. 

  985. <p>#</p>
    986. 

  986. </dd></dl>
    987. 

  987. 

  988. <dl class="py function">
    989. 

  989. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.undo_image_preprocessing">
    990. 

  990. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">undo_image_preprocessing</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">im_tensor</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">numpy.ndarray</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#undo_image_preprocessing"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.undo_image_preprocessing" title="Permalink to this definition"></a></dt>
    991. 

  991. <dd><dl class="field-list simple">
    992. 

  992. <dt class="field-odd">Parameters</dt>
    993. 

  993. <dd class="field-odd"><p><strong>im_tensor</strong> – images in a batch after preprocessing for inference, RGB, (B, C, H, W)</p>
    994. 

  994. </dd>
    995. 

  995. <dt class="field-even">Returns</dt>
    996. 

  996. <dd class="field-even"><p>images in a batch in cv2 format, BGR, (B, H, W, C)</p>
    997. 

  997. </dd>
    998. 

  998. </dl>
    999. 

  999. </dd></dl>
    1000. 

  1000. 

  1001. <dl class="py class">
    1002. 

  1002. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.DetectionVisualization">
    1003. 

  1003. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">DetectionVisualization</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#DetectionVisualization"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.DetectionVisualization" title="Permalink to this definition"></a></dt>
    1004. 

  1004. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1005. 

  1005. <dl class="py method">
    1006. 

  1006. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.DetectionVisualization.visualize_batch">
    1007. 

  1007. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">visualize_batch</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">image_tensor:</span> <span class="pre">torch.Tensor,</span> <span class="pre">pred_boxes:</span> <span class="pre">List[torch.Tensor],</span> <span class="pre">target_boxes:</span> <span class="pre">torch.Tensor,</span> <span class="pre">batch_name:</span> <span class="pre">Union[int,</span> <span class="pre">str],</span> <span class="pre">class_names:</span> <span class="pre">List[str],</span> <span class="pre">checkpoint_dir:</span> <span class="pre">Optional[str]</span> <span class="pre">=</span> <span class="pre">None,</span> <span class="pre">undo_preprocessing_func:</span> <span class="pre">Callable[[torch.Tensor],</span> <span class="pre">numpy.ndarray]</span> <span class="pre">=</span> <span class="pre">&lt;function</span> <span class="pre">undo_image_preprocessing&gt;,</span> <span class="pre">box_thickness:</span> <span class="pre">int</span> <span class="pre">=</span> <span class="pre">2,</span> <span class="pre">image_scale:</span> <span class="pre">float</span> <span class="pre">=</span> <span class="pre">1.0,</span> <span class="pre">gt_alpha:</span> <span class="pre">float</span> <span class="pre">=</span> <span class="pre">0.4</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#DetectionVisualization.visualize_batch"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.DetectionVisualization.visualize_batch" title="Permalink to this definition"></a></dt>
    1008. 

  1008. <dd><p>A helper function to visualize detections predicted by a network:
    1009. 

  1009. saves images into a given path with a name that is {batch_name}_{imade_idx_in_the_batch}.jpg, one batch per call.
    1010. 

  1010. Colors are generated on the fly: uniformly sampled from color wheel to support all given classes.</p>
    1011. 

  1011. <dl class="simple">
    1012. 

  1012. <dt>Adjustable:</dt><dd><ul class="simple">
    1013. 

  1013. <li><p>Ground truth box transparency;</p></li>
    1014. 

  1014. <li><p>Box width;</p></li>
    1015. 

  1015. <li><p>Image size (larger or smaller than what’s provided)</p></li>
    1016. 

  1016. </ul>
    1017. 

  1017. </dd>
    1018. 

  1018. </dl>
    1019. 

  1019. <dl class="field-list simple">
    1020. 

  1020. <dt class="field-odd">Parameters</dt>
    1021. 

  1021. <dd class="field-odd"><ul class="simple">
    1022. 

  1022. <li><p><strong>image_tensor</strong> – rgb images, (B, H, W, 3)</p></li>
    1023. 

  1023. <li><p><strong>pred_boxes</strong> – boxes after NMS for each image in a batch, each (Num_boxes, 6),
    1024. 

  1024. values on dim 1 are: x1, y1, x2, y2, confidence, class</p></li>
    1025. 

  1025. <li><p><strong>target_boxes</strong> – (Num_targets, 6), values on dim 1 are: image id in a batch, class, x y w h
    1026. 

  1026. (coordinates scaled to [0, 1])</p></li>
    1027. 

  1027. <li><p><strong>batch_name</strong> – id of the current batch to use for image naming</p></li>
    1028. 

  1028. <li><p><strong>class_names</strong> – names of all classes, each on its own index</p></li>
    1029. 

  1029. <li><p><strong>checkpoint_dir</strong> – a path where images with boxes will be saved. if None, the result images will
    1030. 

  1030. be returns as a list of numpy image arrays</p></li>
    1031. 

  1031. <li><p><strong>undo_preprocessing_func</strong> – a function to convert preprocessed images tensor into a batch of cv2-like images</p></li>
    1032. 

  1032. <li><p><strong>box_thickness</strong> – box line thickness in px</p></li>
    1033. 

  1033. <li><p><strong>image_scale</strong> – scale of an image w.r.t. given image size,
    1034. 

  1034. e.g. incoming images are (320x320), use scale = 2. to preview in (640x640)</p></li>
    1035. 

  1035. <li><p><strong>gt_alpha</strong> – a value in [0., 1.] transparency on ground truth boxes,
    1036. 

  1036. 0 for invisible, 1 for fully opaque</p></li>
    1037. 

  1037. </ul>
    1038. 

  1038. </dd>
    1039. 

  1039. </dl>
    1040. 

  1040. </dd></dl>
    1041. 

  1041. 

  1042. </dd></dl>
    1043. 

  1043. 

  1044. <dl class="py class">
    1045. 

  1045. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors">
    1046. 

  1046. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.detection_utils.</span></span><span class="sig-name descname"><span class="pre">Anchors</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">anchors_list</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">strides</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/detection_utils.html#Anchors"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors" title="Permalink to this definition"></a></dt>
    1047. 

  1047. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1048. 

  1048. <p>A wrapper function to hold the anchors used by detection models such as Yolo</p>
    1049. 

  1049. <dl class="py property">
    1050. 

  1050. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.stride">
    1051. 

  1051. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">stride</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.nn.parameter.Parameter</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.stride" title="Permalink to this definition"></a></dt>
    1052. 

  1052. <dd></dd></dl>
    1053. 

  1053. 

  1054. <dl class="py property">
    1055. 

  1055. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.anchors">
    1056. 

  1056. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">anchors</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.nn.parameter.Parameter</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.anchors" title="Permalink to this definition"></a></dt>
    1057. 

  1057. <dd></dd></dl>
    1058. 

  1058. 

  1059. <dl class="py property">
    1060. 

  1060. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.anchor_grid">
    1061. 

  1061. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">anchor_grid</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.nn.parameter.Parameter</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.anchor_grid" title="Permalink to this definition"></a></dt>
    1062. 

  1062. <dd></dd></dl>
    1063. 

  1063. 

  1064. <dl class="py property">
    1065. 

  1065. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.detection_layers_num">
    1066. 

  1066. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">detection_layers_num</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.detection_layers_num" title="Permalink to this definition"></a></dt>
    1067. 

  1067. <dd></dd></dl>
    1068. 

  1068. 

  1069. <dl class="py property">
    1070. 

  1070. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.num_anchors">
    1071. 

  1071. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">num_anchors</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.num_anchors" title="Permalink to this definition"></a></dt>
    1072. 

  1072. <dd></dd></dl>
    1073. 

  1073. 

  1074. <dl class="py attribute">
    1075. 

  1075. <dt class="sig sig-object py" id="super_gradients.training.utils.detection_utils.Anchors.training">
    1076. 

  1076. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.detection_utils.Anchors.training" title="Permalink to this definition"></a></dt>
    1077. 

  1077. <dd></dd></dl>
    1078. 

  1078. 

  1079. </dd></dl>
    1080. 

  1080. 

  1081. </section>
    1082. 

  1082. <section id="module-super_gradients.training.utils.distributed_training_utils">
    1083. 

  1083. <span id="super-gradients-training-utils-distributed-training-utils-module"></span><h2>super_gradients.training.utils.distributed_training_utils module<a class="headerlink" href="#module-super_gradients.training.utils.distributed_training_utils" title="Permalink to this headline"></a></h2>
    1084. 

  1084. <dl class="py function">
    1085. 

  1085. <dt class="sig sig-object py" id="super_gradients.training.utils.distributed_training_utils.distributed_all_reduce_tensor_average">
    1086. 

  1086. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.distributed_training_utils.</span></span><span class="sig-name descname"><span class="pre">distributed_all_reduce_tensor_average</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">n</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/distributed_training_utils.html#distributed_all_reduce_tensor_average"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.distributed_training_utils.distributed_all_reduce_tensor_average" title="Permalink to this definition"></a></dt>
    1087. 

  1087. <dd><p>This method performs a reduce operation on multiple nodes running distributed training
    1088. 

  1088. It first sums all of the results and then divides the summation
    1089. 

  1089. :param tensor:  The tensor to perform the reduce operation for
    1090. 

  1090. :param n:  Number of nodes
    1091. 

  1091. :return:   Averaged tensor from all of the nodes</p>
    1092. 

  1092. </dd></dl>
    1093. 

  1093. 

  1094. <dl class="py function">
    1095. 

  1095. <dt class="sig sig-object py" id="super_gradients.training.utils.distributed_training_utils.reduce_results_tuple_for_ddp">
    1096. 

  1096. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.distributed_training_utils.</span></span><span class="sig-name descname"><span class="pre">reduce_results_tuple_for_ddp</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">validation_results_tuple</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/distributed_training_utils.html#reduce_results_tuple_for_ddp"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.distributed_training_utils.reduce_results_tuple_for_ddp" title="Permalink to this definition"></a></dt>
    1097. 

  1097. <dd><p>Gather all validation tuples from the various devices and average them</p>
    1098. 

  1098. </dd></dl>
    1099. 

  1099. 

  1100. <dl class="py class">
    1101. 

  1101. <dt class="sig sig-object py" id="super_gradients.training.utils.distributed_training_utils.MultiGPUModeAutocastWrapper">
    1102. 

  1102. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.distributed_training_utils.</span></span><span class="sig-name descname"><span class="pre">MultiGPUModeAutocastWrapper</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">func</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/distributed_training_utils.html#MultiGPUModeAutocastWrapper"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.distributed_training_utils.MultiGPUModeAutocastWrapper" title="Permalink to this definition"></a></dt>
    1103. 

  1103. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1104. 

  1104. </dd></dl>
    1105. 

  1105. 

  1106. <dl class="py function">
    1107. 

  1107. <dt class="sig sig-object py" id="super_gradients.training.utils.distributed_training_utils.scaled_all_reduce">
    1108. 

  1108. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.distributed_training_utils.</span></span><span class="sig-name descname"><span class="pre">scaled_all_reduce</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tensors</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_gpus</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/distributed_training_utils.html#scaled_all_reduce"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.distributed_training_utils.scaled_all_reduce" title="Permalink to this definition"></a></dt>
    1109. 

  1109. <dd><p>Performs the scaled all_reduce operation on the provided tensors.
    1110. 

  1110. The input tensors are modified in-place.
    1111. 

  1111. Currently supports only the sum
    1112. 

  1112. reduction operator.
    1113. 

  1113. The reduced values are scaled by the inverse size of the
    1114. 

  1114. process group (equivalent to num_gpus).</p>
    1115. 

  1115. </dd></dl>
    1116. 

  1116. 

  1117. <dl class="py function">
    1118. 

  1118. <dt class="sig sig-object py" id="super_gradients.training.utils.distributed_training_utils.compute_precise_bn_stats">
    1119. 

  1119. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.distributed_training_utils.</span></span><span class="sig-name descname"><span class="pre">compute_precise_bn_stats</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">loader</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.utils.data.dataloader.DataLoader</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">precise_bn_batch_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_gpus</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/distributed_training_utils.html#compute_precise_bn_stats"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.distributed_training_utils.compute_precise_bn_stats" title="Permalink to this definition"></a></dt>
    1120. 

  1120. <dd><dl class="field-list simple">
    1121. 

  1121. <dt class="field-odd">Parameters</dt>
    1122. 

  1122. <dd class="field-odd"><ul class="simple">
    1123. 

  1123. <li><p><strong>model</strong> – The model being trained (ie: SgModel.net)</p></li>
    1124. 

  1124. <li><p><strong>loader</strong> – Training dataloader (ie: SgModel.train_loader)</p></li>
    1125. 

  1125. <li><p><strong>precise_bn_batch_size</strong> – The effective batch size we want to calculate the batchnorm on. For example, if we are training a model
    1126. 

  1126. on 8 gpus, with a batch of 128 on each gpu, a good rule of thumb would be to give it 8192
    1127. 

  1127. (ie: effective_batch_size * num_gpus = batch_per_gpu * num_gpus * num_gpus).
    1128. 

  1128. If precise_bn_batch_size is not provided in the training_params, the latter heuristic
    1129. 

  1129. will be taken.</p></li>
    1130. 

  1130. </ul>
    1131. 

  1131. </dd>
    1132. 

  1132. </dl>
    1133. 

  1133. <p>param num_gpus:                 The number of gpus we are training on</p>
    1134. 

  1134. </dd></dl>
    1135. 

  1135. 

  1136. </section>
    1137. 

  1137. <section id="module-super_gradients.training.utils.early_stopping">
    1138. 

  1138. <span id="super-gradients-training-utils-early-stopping-module"></span><h2>super_gradients.training.utils.early_stopping module<a class="headerlink" href="#module-super_gradients.training.utils.early_stopping" title="Permalink to this headline"></a></h2>
    1139. 

  1139. <dl class="py class">
    1140. 

  1140. <dt class="sig sig-object py" id="super_gradients.training.utils.early_stopping.EarlyStop">
    1141. 

  1141. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.early_stopping.</span></span><span class="sig-name descname"><span class="pre">EarlyStop</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">phase</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="#super_gradients.training.utils.callbacks.Phase" title="super_gradients.training.utils.callbacks.Phase"><span class="pre">super_gradients.training.utils.callbacks.Phase</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">monitor</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mode</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'min'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">min_delta</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">patience</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">3</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">check_finite</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">verbose</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">strict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/early_stopping.html#EarlyStop"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.early_stopping.EarlyStop" title="Permalink to this definition"></a></dt>
    1142. 

  1142. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.callbacks.PhaseCallback" title="super_gradients.training.utils.callbacks.PhaseCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.callbacks.PhaseCallback</span></code></a></p>
    1143. 

  1143. <p>Callback to monitor a metric and stop training when it stops improving.
    1144. 

  1144. Inspired by pytorch_lightning.callbacks.early_stopping and tf.keras.callbacks.EarlyStopping</p>
    1145. 

  1145. <dl class="py attribute">
    1146. 

  1146. <dt class="sig sig-object py" id="super_gradients.training.utils.early_stopping.EarlyStop.mode_dict">
    1147. 

  1147. <span class="sig-name descname"><span class="pre">mode_dict</span></span><em class="property"> <span class="pre">=</span> <span class="pre">{'max':</span> <span class="pre">&lt;built-in</span> <span class="pre">method</span> <span class="pre">gt</span> <span class="pre">of</span> <span class="pre">type</span> <span class="pre">object&gt;,</span> <span class="pre">'min':</span> <span class="pre">&lt;built-in</span> <span class="pre">method</span> <span class="pre">lt</span> <span class="pre">of</span> <span class="pre">type</span> <span class="pre">object&gt;}</span></em><a class="headerlink" href="#super_gradients.training.utils.early_stopping.EarlyStop.mode_dict" title="Permalink to this definition"></a></dt>
    1148. 

  1148. <dd></dd></dl>
    1149. 

  1149. 

  1150. <dl class="py attribute">
    1151. 

  1151. <dt class="sig sig-object py" id="super_gradients.training.utils.early_stopping.EarlyStop.supported_phases">
    1152. 

  1152. <span class="sig-name descname"><span class="pre">supported_phases</span></span><em class="property"> <span class="pre">=</span> <span class="pre">(&lt;Phase.VALIDATION_EPOCH_END:</span> <span class="pre">'VALIDATION_EPOCH_END'&gt;,</span> <span class="pre">&lt;Phase.TRAIN_EPOCH_END:</span> <span class="pre">'TRAIN_EPOCH_END'&gt;)</span></em><a class="headerlink" href="#super_gradients.training.utils.early_stopping.EarlyStop.supported_phases" title="Permalink to this definition"></a></dt>
    1153. 

  1153. <dd></dd></dl>
    1154. 

  1154. 

  1155. </dd></dl>
    1156. 

  1156. 

  1157. <dl class="py exception">
    1158. 

  1158. <dt class="sig sig-object py" id="super_gradients.training.utils.early_stopping.MissingMonitorKeyException">
    1159. 

  1159. <em class="property"><span class="pre">exception</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.early_stopping.</span></span><span class="sig-name descname"><span class="pre">MissingMonitorKeyException</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/early_stopping.html#MissingMonitorKeyException"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.early_stopping.MissingMonitorKeyException" title="Permalink to this definition"></a></dt>
    1160. 

  1160. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Exception</span></code></p>
    1161. 

  1161. <p>Exception raised for missing monitor key in metrics_dict.</p>
    1162. 

  1162. </dd></dl>
    1163. 

  1163. 

  1164. </section>
    1165. 

  1165. <section id="module-super_gradients.training.utils.ema">
    1166. 

  1166. <span id="super-gradients-training-utils-ema-module"></span><h2>super_gradients.training.utils.ema module<a class="headerlink" href="#module-super_gradients.training.utils.ema" title="Permalink to this headline"></a></h2>
    1167. 

  1167. <dl class="py function">
    1168. 

  1168. <dt class="sig sig-object py" id="super_gradients.training.utils.ema.copy_attr">
    1169. 

  1169. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.ema.</span></span><span class="sig-name descname"><span class="pre">copy_attr</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">a</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">b</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">include</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">()</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exclude</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">()</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ema.html#copy_attr"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ema.copy_attr" title="Permalink to this definition"></a></dt>
    1170. 

  1170. <dd></dd></dl>
    1171. 

  1171. 

  1172. <dl class="py class">
    1173. 

  1173. <dt class="sig sig-object py" id="super_gradients.training.utils.ema.ModelEMA">
    1174. 

  1174. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.ema.</span></span><span class="sig-name descname"><span class="pre">ModelEMA</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">decay</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.9999</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">beta</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">15</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exp_activation</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ema.html#ModelEMA"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ema.ModelEMA" title="Permalink to this definition"></a></dt>
    1175. 

  1175. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1176. 

  1176. <p>Model Exponential Moving Average from <a class="reference external" href="https://github.com/rwightman/pytorch-image-models">https://github.com/rwightman/pytorch-image-models</a>
    1177. 

  1177. Keep a moving average of everything in the model state_dict (parameters and buffers).
    1178. 

  1178. This is intended to allow functionality like
    1179. 

  1179. <a class="reference external" href="https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage">https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage</a>
    1180. 

  1180. A smoothed version of the weights is necessary for some training schemes to perform well.
    1181. 

  1181. This class is sensitive where it is initialized in the sequence of model init,
    1182. 

  1182. GPU assignment and distributed training wrappers.</p>
    1183. 

  1183. <dl class="py method">
    1184. 

  1184. <dt class="sig sig-object py" id="super_gradients.training.utils.ema.ModelEMA.update">
    1185. 

  1185. <span class="sig-name descname"><span class="pre">update</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">training_percent</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ema.html#ModelEMA.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ema.ModelEMA.update" title="Permalink to this definition"></a></dt>
    1186. 

  1186. <dd><p>Update the state of the EMA model.
    1187. 

  1187. :param model: current training model
    1188. 

  1188. :param training_percent: the percentage of the training process [0,1]. i.e 0.4 means 40% of the training have passed</p>
    1189. 

  1189. </dd></dl>
    1190. 

  1190. 

  1191. <dl class="py method">
    1192. 

  1192. <dt class="sig sig-object py" id="super_gradients.training.utils.ema.ModelEMA.update_attr">
    1193. 

  1193. <span class="sig-name descname"><span class="pre">update_attr</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ema.html#ModelEMA.update_attr"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ema.ModelEMA.update_attr" title="Permalink to this definition"></a></dt>
    1194. 

  1194. <dd><p>This function updates model attributes (not weight and biases) from original model to the ema model.
    1195. 

  1195. attributes of the original model, such as anchors and grids (of detection models), may be crucial to the
    1196. 

  1196. model operation and need to be updated.
    1197. 

  1197. If include_attributes and exclude_attributes lists were not defined, all non-private (not starting with ‘_’)
    1198. 

  1198. attributes will be updated (and only them).
    1199. 

  1199. :param model: the source model</p>
    1200. 

  1200. </dd></dl>
    1201. 

  1201. 

  1202. </dd></dl>
    1203. 

  1203. 

  1204. </section>
    1205. 

  1205. <section id="module-super_gradients.training.utils.export_utils">
    1206. 

  1206. <span id="super-gradients-training-utils-export-utils-module"></span><h2>super_gradients.training.utils.export_utils module<a class="headerlink" href="#module-super_gradients.training.utils.export_utils" title="Permalink to this headline"></a></h2>
    1207. 

  1207. <dl class="py class">
    1208. 

  1208. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableHardswish">
    1209. 

  1209. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.export_utils.</span></span><span class="sig-name descname"><span class="pre">ExportableHardswish</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/export_utils.html#ExportableHardswish"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableHardswish" title="Permalink to this definition"></a></dt>
    1210. 

  1210. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1211. 

  1211. <p>Export-friendly version of nn.Hardswish()</p>
    1212. 

  1212. <dl class="py method">
    1213. 

  1213. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableHardswish.forward">
    1214. 

  1214. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/export_utils.html#ExportableHardswish.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableHardswish.forward" title="Permalink to this definition"></a></dt>
    1215. 

  1215. <dd><p>Defines the computation performed at every call.</p>
    1216. 

  1216. <p>Should be overridden by all subclasses.</p>
    1217. 

  1217. <div class="admonition note">
    1218. 

  1218. <p class="admonition-title">Note</p>
    1219. 

  1219. <p>Although the recipe for forward pass needs to be defined within
    1220. 

  1220. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1221. 

  1221. instead of this since the former takes care of running the
    1222. 

  1222. registered hooks while the latter silently ignores them.</p>
    1223. 

  1223. </div>
    1224. 

  1224. </dd></dl>
    1225. 

  1225. 

  1226. <dl class="py attribute">
    1227. 

  1227. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableHardswish.training">
    1228. 

  1228. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableHardswish.training" title="Permalink to this definition"></a></dt>
    1229. 

  1229. <dd></dd></dl>
    1230. 

  1230. 

  1231. </dd></dl>
    1232. 

  1232. 

  1233. <dl class="py class">
    1234. 

  1234. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableSiLU">
    1235. 

  1235. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.export_utils.</span></span><span class="sig-name descname"><span class="pre">ExportableSiLU</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/export_utils.html#ExportableSiLU"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableSiLU" title="Permalink to this definition"></a></dt>
    1236. 

  1236. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1237. 

  1237. <p>Export-friendly version of nn.SiLU()
    1238. 

  1238. From <a class="reference external" href="https://github.com/ultralytics/yolov5">https://github.com/ultralytics/yolov5</a></p>
    1239. 

  1239. <dl class="py method">
    1240. 

  1240. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableSiLU.forward">
    1241. 

  1241. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/export_utils.html#ExportableSiLU.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableSiLU.forward" title="Permalink to this definition"></a></dt>
    1242. 

  1242. <dd><p>Defines the computation performed at every call.</p>
    1243. 

  1243. <p>Should be overridden by all subclasses.</p>
    1244. 

  1244. <div class="admonition note">
    1245. 

  1245. <p class="admonition-title">Note</p>
    1246. 

  1246. <p>Although the recipe for forward pass needs to be defined within
    1247. 

  1247. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1248. 

  1248. instead of this since the former takes care of running the
    1249. 

  1249. registered hooks while the latter silently ignores them.</p>
    1250. 

  1250. </div>
    1251. 

  1251. </dd></dl>
    1252. 

  1252. 

  1253. <dl class="py attribute">
    1254. 

  1254. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.ExportableSiLU.training">
    1255. 

  1255. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.export_utils.ExportableSiLU.training" title="Permalink to this definition"></a></dt>
    1256. 

  1256. <dd></dd></dl>
    1257. 

  1257. 

  1258. </dd></dl>
    1259. 

  1259. 

  1260. <dl class="py function">
    1261. 

  1261. <dt class="sig sig-object py" id="super_gradients.training.utils.export_utils.fuse_conv_bn">
    1262. 

  1262. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.export_utils.</span></span><span class="sig-name descname"><span class="pre">fuse_conv_bn</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">replace_bn_with_identity</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/export_utils.html#fuse_conv_bn"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.export_utils.fuse_conv_bn" title="Permalink to this definition"></a></dt>
    1263. 

  1263. <dd><p>Fuses consecutive nn.Conv2d and nn.BatchNorm2d layers recursively inplace in all of the model
    1264. 

  1264. :param replace_bn_with_identity: if set to true, bn will be replaced with identity. otherwise, bn will be removed
    1265. 

  1265. :param model: the target model
    1266. 

  1266. :return: the number of fuses executed</p>
    1267. 

  1267. </dd></dl>
    1268. 

  1268. 

  1269. </section>
    1270. 

  1270. <section id="super-gradients-training-utils-get-model-stats-module">
    1271. 

  1271. <h2>super_gradients.training.utils.get_model_stats module<a class="headerlink" href="#super-gradients-training-utils-get-model-stats-module" title="Permalink to this headline"></a></h2>
    1272. 

  1272. </section>
    1273. 

  1273. <section id="module-super_gradients.training.utils.module_utils">
    1274. 

  1274. <span id="super-gradients-training-utils-module-utils-module"></span><h2>super_gradients.training.utils.module_utils module<a class="headerlink" href="#module-super_gradients.training.utils.module_utils" title="Permalink to this headline"></a></h2>
    1275. 

  1275. <dl class="py class">
    1276. 

  1276. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.MultiOutputModule">
    1277. 

  1277. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.module_utils.</span></span><span class="sig-name descname"><span class="pre">MultiOutputModule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">module</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">output_paths</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prune</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#MultiOutputModule"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.MultiOutputModule" title="Permalink to this definition"></a></dt>
    1278. 

  1278. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1279. 

  1279. <p>This module wraps around a container nn.Module (such as Module, Sequential and ModuleList) and allows to extract
    1280. 

  1280. multiple output from its inner modules on each forward call() (as a list of output tensors)
    1281. 

  1281. note: the default output of the wrapped module will not be added to the output list by default. To get
    1282. 

  1282. the default output in the outputs list, explicitly include its path in the &#64;output_paths parameter</p>
    1283. 

  1283. <p>i.e.
    1284. 

  1284. for module:</p>
    1285. 

  1285. <blockquote>
    1286. 

  1286. <div><dl>
    1287. 

  1287. <dt>Sequential(</dt><dd><dl class="simple">
    1288. 

  1288. <dt>(0): Sequential(</dt><dd><p>(0): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
    1289. 

  1289. (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    1290. 

  1290. (2): ReLU6(inplace=True)</p>
    1291. 

  1291. </dd>
    1292. 

  1292. </dl>
    1293. 

  1293. <p>)                                         ===================================&gt;&gt;
    1294. 

  1294. (1): InvertedResidual(</p>
    1295. 

  1295. <blockquote>
    1296. 

  1296. <div><dl class="simple">
    1297. 

  1297. <dt>(conv): Sequential(</dt><dd><p>(0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
    1298. 

  1298. (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    1299. 

  1299. (2): ReLU6(inplace=True)              ===================================&gt;&gt;
    1300. 

  1300. (3): Conv2d(32, 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
    1301. 

  1301. (4): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)</p>
    1302. 

  1302. </dd>
    1303. 

  1303. </dl>
    1304. 

  1304. <p>)</p>
    1305. 

  1305. </div></blockquote>
    1306. 

  1306. <p>)</p>
    1307. 

  1307. </dd>
    1308. 

  1308. </dl>
    1309. 

  1309. <p>)</p>
    1310. 

  1310. </div></blockquote>
    1311. 

  1311. <dl class="simple">
    1312. 

  1312. <dt>and paths:</dt><dd><p>[0, [1, ‘conv’, 2]]</p>
    1313. 

  1313. </dd>
    1314. 

  1314. </dl>
    1315. 

  1315. <p>the output are marked with arrows</p>
    1316. 

  1316. <dl class="py method">
    1317. 

  1317. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.MultiOutputModule.save_output_hook">
    1318. 

  1318. <span class="sig-name descname"><span class="pre">save_output_hook</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">_</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">input</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">output</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#MultiOutputModule.save_output_hook"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.MultiOutputModule.save_output_hook" title="Permalink to this definition"></a></dt>
    1319. 

  1319. <dd></dd></dl>
    1320. 

  1320. 

  1321. <dl class="py method">
    1322. 

  1322. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.MultiOutputModule.forward">
    1323. 

  1323. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">list</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#MultiOutputModule.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.MultiOutputModule.forward" title="Permalink to this definition"></a></dt>
    1324. 

  1324. <dd><p>Defines the computation performed at every call.</p>
    1325. 

  1325. <p>Should be overridden by all subclasses.</p>
    1326. 

  1326. <div class="admonition note">
    1327. 

  1327. <p class="admonition-title">Note</p>
    1328. 

  1328. <p>Although the recipe for forward pass needs to be defined within
    1329. 

  1329. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1330. 

  1330. instead of this since the former takes care of running the
    1331. 

  1331. registered hooks while the latter silently ignores them.</p>
    1332. 

  1332. </div>
    1333. 

  1333. </dd></dl>
    1334. 

  1334. 

  1335. <dl class="py attribute">
    1336. 

  1336. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.MultiOutputModule.training">
    1337. 

  1337. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.module_utils.MultiOutputModule.training" title="Permalink to this definition"></a></dt>
    1338. 

  1338. <dd></dd></dl>
    1339. 

  1339. 

  1340. </dd></dl>
    1341. 

  1341. 

  1342. <dl class="py function">
    1343. 

  1343. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.replace_activations">
    1344. 

  1344. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.module_utils.</span></span><span class="sig-name descname"><span class="pre">replace_activations</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">module</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">new_activation</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">activations_to_replace</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">type</span><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#replace_activations"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.replace_activations" title="Permalink to this definition"></a></dt>
    1345. 

  1345. <dd><p>Recursively go through module and replaces each activation in activations_to_replace with a copy of new_activation
    1346. 

  1346. :param module:                  a module that will be changed inplace
    1347. 

  1347. :param new_activation:          a sample of a new activation (will be copied)
    1348. 

  1348. :param activations_to_replace:  types of activations to replace, each must be a subclass of nn.Module</p>
    1349. 

  1349. </dd></dl>
    1350. 

  1350. 

  1351. <dl class="py function">
    1352. 

  1352. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.fuse_repvgg_blocks_residual_branches">
    1353. 

  1353. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.module_utils.</span></span><span class="sig-name descname"><span class="pre">fuse_repvgg_blocks_residual_branches</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#fuse_repvgg_blocks_residual_branches"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.fuse_repvgg_blocks_residual_branches" title="Permalink to this definition"></a></dt>
    1354. 

  1354. <dd><p>Call fuse_block_residual_branches for all repvgg blocks in the model
    1355. 

  1355. :param model: torch.nn.Module with repvgg blocks. Doesn’t have to be entirely consists of repvgg.
    1356. 

  1356. :type model: torch.nn.Module</p>
    1357. 

  1357. </dd></dl>
    1358. 

  1358. 

  1359. <dl class="py class">
    1360. 

  1360. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.ConvBNReLU">
    1361. 

  1361. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.module_utils.</span></span><span class="sig-name descname"><span class="pre">ConvBNReLU</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">in_channels</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">out_channels</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">kernel_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">int</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stride</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">int</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">padding</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">int</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dilation</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">int</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">groups</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">bias</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">padding_mode</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'zeros'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">use_normalization</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1e-05</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">momentum</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">affine</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">track_running_stats</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dtype</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">use_activation</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">inplace</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#ConvBNReLU"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.ConvBNReLU" title="Permalink to this definition"></a></dt>
    1362. 

  1362. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1363. 

  1363. <dl class="simple">
    1364. 

  1364. <dt>Class for Convolution2d-Batchnorm2d-Relu layer. Default behaviour is Conv-BN-Relu. To exclude Batchnorm module use</dt><dd><p><cite>use_normalization=False</cite>, to exclude Relu activation use <cite>use_activation=False</cite>.</p>
    1365. 

  1365. </dd>
    1366. 

  1366. </dl>
    1367. 

  1367. <p>For convolution arguments documentation see <cite>nn.Conv2d</cite>.
    1368. 

  1368. For batchnorm arguments documentation see <cite>nn.BatchNorm2d</cite>.
    1369. 

  1369. For relu arguments documentation see <cite>nn.Relu</cite>.</p>
    1370. 

  1370. <dl class="py method">
    1371. 

  1371. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.ConvBNReLU.forward">
    1372. 

  1372. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/module_utils.html#ConvBNReLU.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.module_utils.ConvBNReLU.forward" title="Permalink to this definition"></a></dt>
    1373. 

  1373. <dd><p>Defines the computation performed at every call.</p>
    1374. 

  1374. <p>Should be overridden by all subclasses.</p>
    1375. 

  1375. <div class="admonition note">
    1376. 

  1376. <p class="admonition-title">Note</p>
    1377. 

  1377. <p>Although the recipe for forward pass needs to be defined within
    1378. 

  1378. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1379. 

  1379. instead of this since the former takes care of running the
    1380. 

  1380. registered hooks while the latter silently ignores them.</p>
    1381. 

  1381. </div>
    1382. 

  1382. </dd></dl>
    1383. 

  1383. 

  1384. <dl class="py attribute">
    1385. 

  1385. <dt class="sig sig-object py" id="super_gradients.training.utils.module_utils.ConvBNReLU.training">
    1386. 

  1386. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.module_utils.ConvBNReLU.training" title="Permalink to this definition"></a></dt>
    1387. 

  1387. <dd></dd></dl>
    1388. 

  1388. 

  1389. </dd></dl>
    1390. 

  1390. 

  1391. </section>
    1392. 

  1392. <section id="module-super_gradients.training.utils.optimizer_utils">
    1393. 

  1393. <span id="super-gradients-training-utils-optimizer-utils-module"></span><h2>super_gradients.training.utils.optimizer_utils module<a class="headerlink" href="#module-super_gradients.training.utils.optimizer_utils" title="Permalink to this headline"></a></h2>
    1394. 

  1394. <dl class="py function">
    1395. 

  1395. <dt class="sig sig-object py" id="super_gradients.training.utils.optimizer_utils.separate_zero_wd_params_groups_for_optimizer">
    1396. 

  1396. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.optimizer_utils.</span></span><span class="sig-name descname"><span class="pre">separate_zero_wd_params_groups_for_optimizer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">module</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">net_named_params</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight_decay</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/optimizer_utils.html#separate_zero_wd_params_groups_for_optimizer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.optimizer_utils.separate_zero_wd_params_groups_for_optimizer" title="Permalink to this definition"></a></dt>
    1397. 

  1397. <dd><dl class="simple">
    1398. 

  1398. <dt>separate param groups for batchnorm and biases and others with weight decay. return list of param groups in format</dt><dd><p>required by torch Optimizer classes.</p>
    1399. 

  1399. </dd>
    1400. 

  1400. <dt>bias + BN with weight decay=0 and the rest with the given weight decay</dt><dd><dl class="field-list simple">
    1401. 

  1401. <dt class="field-odd">param module</dt>
    1402. 

  1402. <dd class="field-odd"><p>train net module.</p>
    1403. 

  1403. </dd>
    1404. 

  1404. <dt class="field-even">param net_named_params</dt>
    1405. 

  1405. <dd class="field-even"><p>list of params groups, output of SgModule.initialize_param_groups</p>
    1406. 

  1406. </dd>
    1407. 

  1407. <dt class="field-odd">param weight_decay</dt>
    1408. 

  1408. <dd class="field-odd"><p>value to set for the non BN and bias parameters</p>
    1409. 

  1409. </dd>
    1410. 

  1410. </dl>
    1411. 

  1411. </dd>
    1412. 

  1412. </dl>
    1413. 

  1413. </dd></dl>
    1414. 

  1414. 

  1415. <dl class="py function">
    1416. 

  1416. <dt class="sig sig-object py" id="super_gradients.training.utils.optimizer_utils.build_optimizer">
    1417. 

  1417. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.optimizer_utils.</span></span><span class="sig-name descname"><span class="pre">build_optimizer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">net</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">training_params</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/optimizer_utils.html#build_optimizer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.optimizer_utils.build_optimizer" title="Permalink to this definition"></a></dt>
    1418. 

  1418. <dd><dl class="simple">
    1419. 

  1419. <dt>Wrapper function for initializing the optimizer</dt><dd><dl class="field-list simple">
    1420. 

  1420. <dt class="field-odd">param net</dt>
    1421. 

  1421. <dd class="field-odd"><p>the nn_module to build the optimizer for</p>
    1422. 

  1422. </dd>
    1423. 

  1423. <dt class="field-even">param lr</dt>
    1424. 

  1424. <dd class="field-even"><p>initial learning rate</p>
    1425. 

  1425. </dd>
    1426. 

  1426. <dt class="field-odd">param training_params</dt>
    1427. 

  1427. <dd class="field-odd"><p>training_parameters</p>
    1428. 

  1428. </dd>
    1429. 

  1429. </dl>
    1430. 

  1430. </dd>
    1431. 

  1431. </dl>
    1432. 

  1432. </dd></dl>
    1433. 

  1433. 

  1434. </section>
    1435. 

  1435. <section id="module-super_gradients.training.utils.regularization_utils">
    1436. 

  1436. <span id="super-gradients-training-utils-regularization-utils-module"></span><h2>super_gradients.training.utils.regularization_utils module<a class="headerlink" href="#module-super_gradients.training.utils.regularization_utils" title="Permalink to this headline"></a></h2>
    1437. 

  1437. <dl class="py class">
    1438. 

  1438. <dt class="sig sig-object py" id="super_gradients.training.utils.regularization_utils.DropPath">
    1439. 

  1439. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.regularization_utils.</span></span><span class="sig-name descname"><span class="pre">DropPath</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">drop_prob</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/regularization_utils.html#DropPath"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.regularization_utils.DropPath" title="Permalink to this definition"></a></dt>
    1440. 

  1440. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1441. 

  1441. <p>Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks).</p>
    1442. 

  1442. <p>Code taken from TIMM (<a class="reference external" href="https://github.com/rwightman/pytorch-image-models">https://github.com/rwightman/pytorch-image-models</a>)
    1443. 

  1443. Apache License 2.0</p>
    1444. 

  1444. <dl class="py method">
    1445. 

  1445. <dt class="sig sig-object py" id="super_gradients.training.utils.regularization_utils.DropPath.forward">
    1446. 

  1446. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/regularization_utils.html#DropPath.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.regularization_utils.DropPath.forward" title="Permalink to this definition"></a></dt>
    1447. 

  1447. <dd><p>Defines the computation performed at every call.</p>
    1448. 

  1448. <p>Should be overridden by all subclasses.</p>
    1449. 

  1449. <div class="admonition note">
    1450. 

  1450. <p class="admonition-title">Note</p>
    1451. 

  1451. <p>Although the recipe for forward pass needs to be defined within
    1452. 

  1452. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1453. 

  1453. instead of this since the former takes care of running the
    1454. 

  1454. registered hooks while the latter silently ignores them.</p>
    1455. 

  1455. </div>
    1456. 

  1456. </dd></dl>
    1457. 

  1457. 

  1458. <dl class="py attribute">
    1459. 

  1459. <dt class="sig sig-object py" id="super_gradients.training.utils.regularization_utils.DropPath.training">
    1460. 

  1460. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.regularization_utils.DropPath.training" title="Permalink to this definition"></a></dt>
    1461. 

  1461. <dd></dd></dl>
    1462. 

  1462. 

  1463. </dd></dl>
    1464. 

  1464. 

  1465. </section>
    1466. 

  1466. <section id="module-super_gradients.training.utils.segmentation_utils">
    1467. 

  1467. <span id="super-gradients-training-utils-segmentation-utils-module"></span><h2>super_gradients.training.utils.segmentation_utils module<a class="headerlink" href="#module-super_gradients.training.utils.segmentation_utils" title="Permalink to this headline"></a></h2>
    1468. 

  1468. <dl class="py class">
    1469. 

  1469. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.SegmentationTransform">
    1470. 

  1470. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">SegmentationTransform</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#SegmentationTransform"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="Permalink to this definition"></a></dt>
    1471. 

  1471. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1472. 

  1472. </dd></dl>
    1473. 

  1473. 

  1474. <dl class="py class">
    1475. 

  1475. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomFlip">
    1476. 

  1476. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">RandomFlip</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prob</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomFlip"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomFlip" title="Permalink to this definition"></a></dt>
    1477. 

  1477. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1478. 

  1478. <p>Randomly flips the image and mask (synchronously) with probability ‘prob’.</p>
    1479. 

  1479. </dd></dl>
    1480. 

  1480. 

  1481. <dl class="py class">
    1482. 

  1482. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.Rescale">
    1483. 

  1483. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">Rescale</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">scale_factor</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">short_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">long_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#Rescale"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.Rescale" title="Permalink to this definition"></a></dt>
    1484. 

  1484. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1485. 

  1485. <p>Rescales the image and mask (synchronously) while preserving aspect ratio.
    1486. 

  1486. The rescaling can be done according to scale_factor, short_size or long_size.
    1487. 

  1487. If more than one argument is given, the rescaling mode is determined by this order: scale_factor, then short_size,
    1488. 

  1488. then long_size.</p>
    1489. 

  1489. <dl class="field-list simple">
    1490. 

  1490. <dt class="field-odd">Parameters</dt>
    1491. 

  1491. <dd class="field-odd"><ul class="simple">
    1492. 

  1492. <li><p><strong>scale_factor</strong> – rescaling is done by multiplying input size by scale_factor:
    1493. 

  1493. out_size = (scale_factor * w, scale_factor * h)</p></li>
    1494. 

  1494. <li><p><strong>short_size</strong> – rescaling is done by determining the scale factor by the ratio short_size / min(h, w).</p></li>
    1495. 

  1495. <li><p><strong>long_size</strong> – rescaling is done by determining the scale factor by the ratio long_size / max(h, w).</p></li>
    1496. 

  1496. </ul>
    1497. 

  1497. </dd>
    1498. 

  1498. </dl>
    1499. 

  1499. <dl class="py method">
    1500. 

  1500. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.Rescale.check_valid_arguments">
    1501. 

  1501. <span class="sig-name descname"><span class="pre">check_valid_arguments</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#Rescale.check_valid_arguments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.Rescale.check_valid_arguments" title="Permalink to this definition"></a></dt>
    1502. 

  1502. <dd></dd></dl>
    1503. 

  1503. 

  1504. </dd></dl>
    1505. 

  1505. 

  1506. <dl class="py class">
    1507. 

  1507. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomRescale">
    1508. 

  1508. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">RandomRescale</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">scales</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">(0.5,</span> <span class="pre">2.0)</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomRescale"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomRescale" title="Permalink to this definition"></a></dt>
    1509. 

  1509. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1510. 

  1510. <p>Random rescale the image and mask (synchronously) while preserving aspect ratio.
    1511. 

  1511. Scale factor is randomly picked between scales [min, max]
    1512. 

  1512. :param scales: scale range tuple (min, max), if scales is a float range will be defined as (1, scales) if scales &gt; 1,</p>
    1513. 

  1513. <blockquote>
    1514. 

  1514. <div><p>otherwise (scales, 1). must be a positive number.</p>
    1515. 

  1515. </div></blockquote>
    1516. 

  1516. <dl class="py method">
    1517. 

  1517. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomRescale.check_valid_arguments">
    1518. 

  1518. <span class="sig-name descname"><span class="pre">check_valid_arguments</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomRescale.check_valid_arguments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomRescale.check_valid_arguments" title="Permalink to this definition"></a></dt>
    1519. 

  1519. <dd><p>Check the scale values are valid. if order is wrong, flip the order and return the right scale values.</p>
    1520. 

  1520. </dd></dl>
    1521. 

  1521. 

  1522. </dd></dl>
    1523. 

  1523. 

  1524. <dl class="py class">
    1525. 

  1525. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomRotate">
    1526. 

  1526. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">RandomRotate</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">min_deg</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">-</span> <span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_deg</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fill_mask</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fill_image</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomRotate"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomRotate" title="Permalink to this definition"></a></dt>
    1527. 

  1527. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1528. 

  1528. <p>Randomly rotates image and mask (synchronously) between ‘min_deg’ and ‘max_deg’.</p>
    1529. 

  1529. <dl class="py method">
    1530. 

  1530. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomRotate.check_valid_arguments">
    1531. 

  1531. <span class="sig-name descname"><span class="pre">check_valid_arguments</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomRotate.check_valid_arguments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomRotate.check_valid_arguments" title="Permalink to this definition"></a></dt>
    1532. 

  1532. <dd></dd></dl>
    1533. 

  1533. 

  1534. </dd></dl>
    1535. 

  1535. 

  1536. <dl class="py class">
    1537. 

  1537. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.CropImageAndMask">
    1538. 

  1538. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">CropImageAndMask</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">crop_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mode</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#CropImageAndMask"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.CropImageAndMask" title="Permalink to this definition"></a></dt>
    1539. 

  1539. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1540. 

  1540. <p>Crops image and mask (synchronously).
    1541. 

  1541. In “center” mode a center crop is performed while, in “random” mode the drop will be positioned around</p>
    1542. 

  1542. <blockquote>
    1543. 

  1543. <div><p>random coordinates.</p>
    1544. 

  1544. </div></blockquote>
    1545. 

  1545. <dl class="py method">
    1546. 

  1546. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.CropImageAndMask.check_valid_arguments">
    1547. 

  1547. <span class="sig-name descname"><span class="pre">check_valid_arguments</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#CropImageAndMask.check_valid_arguments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.CropImageAndMask.check_valid_arguments" title="Permalink to this definition"></a></dt>
    1548. 

  1548. <dd></dd></dl>
    1549. 

  1549. 

  1550. </dd></dl>
    1551. 

  1551. 

  1552. <dl class="py class">
    1553. 

  1553. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.RandomGaussianBlur">
    1554. 

  1554. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">RandomGaussianBlur</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prob</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#RandomGaussianBlur"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.RandomGaussianBlur" title="Permalink to this definition"></a></dt>
    1555. 

  1555. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1556. 

  1556. <p>Adds random Gaussian Blur to image with probability ‘prob’.</p>
    1557. 

  1557. </dd></dl>
    1558. 

  1558. 

  1559. <dl class="py class">
    1560. 

  1560. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.PadShortToCropSize">
    1561. 

  1561. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">PadShortToCropSize</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">crop_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fill_mask</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fill_image</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#PadShortToCropSize"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.PadShortToCropSize" title="Permalink to this definition"></a></dt>
    1562. 

  1562. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.segmentation_utils.SegmentationTransform" title="super_gradients.training.utils.segmentation_utils.SegmentationTransform"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.segmentation_utils.SegmentationTransform</span></code></a></p>
    1563. 

  1563. <p>Pads image to ‘crop_size’.
    1564. 

  1564. Should be called only after “Rescale” or “RandomRescale” in augmentations pipeline.</p>
    1565. 

  1565. <dl class="py method">
    1566. 

  1566. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.PadShortToCropSize.check_valid_arguments">
    1567. 

  1567. <span class="sig-name descname"><span class="pre">check_valid_arguments</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#PadShortToCropSize.check_valid_arguments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.PadShortToCropSize.check_valid_arguments" title="Permalink to this definition"></a></dt>
    1568. 

  1568. <dd></dd></dl>
    1569. 

  1569. 

  1570. </dd></dl>
    1571. 

  1571. 

  1572. <dl class="py class">
    1573. 

  1573. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.ColorJitterSeg">
    1574. 

  1574. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">ColorJitterSeg</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">brightness</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">contrast</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">saturation</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">hue</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#ColorJitterSeg"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.ColorJitterSeg" title="Permalink to this definition"></a></dt>
    1575. 

  1575. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchvision.transforms.transforms.ColorJitter</span></code></p>
    1576. 

  1576. <dl class="py attribute">
    1577. 

  1577. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.ColorJitterSeg.training">
    1578. 

  1578. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.ColorJitterSeg.training" title="Permalink to this definition"></a></dt>
    1579. 

  1579. <dd></dd></dl>
    1580. 

  1580. 

  1581. </dd></dl>
    1582. 

  1582. 

  1583. <dl class="py function">
    1584. 

  1584. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.coco_sub_classes_inclusion_tuples_list">
    1585. 

  1585. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">coco_sub_classes_inclusion_tuples_list</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#coco_sub_classes_inclusion_tuples_list"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.coco_sub_classes_inclusion_tuples_list" title="Permalink to this definition"></a></dt>
    1586. 

  1586. <dd></dd></dl>
    1587. 

  1587. 

  1588. <dl class="py function">
    1589. 

  1589. <dt class="sig sig-object py" id="super_gradients.training.utils.segmentation_utils.to_one_hot">
    1590. 

  1590. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.segmentation_utils.</span></span><span class="sig-name descname"><span class="pre">to_one_hot</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">target</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_classes</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/segmentation_utils.html#to_one_hot"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.segmentation_utils.to_one_hot" title="Permalink to this definition"></a></dt>
    1591. 

  1591. <dd><p>Target label to one_hot tensor. labels and ignore_index must be consecutive numbers.
    1592. 

  1592. :param target: Class labels long tensor, with shape [N, H, W]
    1593. 

  1593. :param num_classes: num of classes in datasets excluding ignore label, this is the output channels of the one hot</p>
    1594. 

  1594. <blockquote>
    1595. 

  1595. <div><p>result.</p>
    1596. 

  1596. </div></blockquote>
    1597. 

  1597. <dl class="field-list simple">
    1598. 

  1598. <dt class="field-odd">Returns</dt>
    1599. 

  1599. <dd class="field-odd"><p>one hot tensor with shape [N, num_classes, H, W]</p>
    1600. 

  1600. </dd>
    1601. 

  1601. </dl>
    1602. 

  1602. </dd></dl>
    1603. 

  1603. 

  1604. </section>
    1605. 

  1605. <section id="module-super_gradients.training.utils.sg_model_utils">
    1606. 

  1606. <span id="super-gradients-training-utils-sg-model-utils-module"></span><h2>super_gradients.training.utils.sg_model_utils module<a class="headerlink" href="#module-super_gradients.training.utils.sg_model_utils" title="Permalink to this headline"></a></h2>
    1607. 

  1607. <dl class="py function">
    1608. 

  1608. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.try_port">
    1609. 

  1609. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">try_port</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">port</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#try_port"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.try_port" title="Permalink to this definition"></a></dt>
    1610. 

  1610. <dd><p>try_port - Helper method for tensorboard port binding
    1611. 

  1611. :param port:
    1612. 

  1612. :return:</p>
    1613. 

  1613. </dd></dl>
    1614. 

  1614. 

  1615. <dl class="py function">
    1616. 

  1616. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.launch_tensorboard_process">
    1617. 

  1617. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">launch_tensorboard_process</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">checkpoints_dir_path</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sleep_postpone</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">bool</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">port</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Optional</span><span class="p"><span class="pre">[</span></span><span class="pre">int</span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">multiprocessing.context.Process</span><span class="p"><span class="pre">,</span> </span><span class="pre">int</span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#launch_tensorboard_process"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.launch_tensorboard_process" title="Permalink to this definition"></a></dt>
    1618. 

  1618. <dd><dl>
    1619. 

  1619. <dt>launch_tensorboard_process - Default behavior is to scan all free ports from 6006-6016 and try using them</dt><dd><blockquote>
    1620. 

  1620. <div><p>unless port is defined by the user</p>
    1621. 

  1621. </div></blockquote>
    1622. 

  1622. <dl class="field-list simple">
    1623. 

  1623. <dt class="field-odd">param checkpoints_dir_path</dt>
    1624. 

  1624. <dd class="field-odd"><p></p></dd>
    1625. 

  1625. <dt class="field-even">param sleep_postpone</dt>
    1626. 

  1626. <dd class="field-even"><p></p></dd>
    1627. 

  1627. <dt class="field-odd">param port</dt>
    1628. 

  1628. <dd class="field-odd"><p></p></dd>
    1629. 

  1629. <dt class="field-even">return</dt>
    1630. 

  1630. <dd class="field-even"><p>tuple of tb process, port</p>
    1631. 

  1631. </dd>
    1632. 

  1632. </dl>
    1633. 

  1633. </dd>
    1634. 

  1634. </dl>
    1635. 

  1635. </dd></dl>
    1636. 

  1636. 

  1637. <dl class="py function">
    1638. 

  1638. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.init_summary_writer">
    1639. 

  1639. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">init_summary_writer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">tb_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">checkpoint_loaded</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">user_prompt</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#init_summary_writer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.init_summary_writer" title="Permalink to this definition"></a></dt>
    1640. 

  1640. <dd><p>Remove previous tensorboard files from directory and launch a tensor board process</p>
    1641. 

  1641. </dd></dl>
    1642. 

  1642. 

  1643. <dl class="py function">
    1644. 

  1644. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.add_log_to_file">
    1645. 

  1645. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">add_log_to_file</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">filename</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">results_titles_list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">results_values_list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">epoch</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_epochs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#add_log_to_file"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.add_log_to_file" title="Permalink to this definition"></a></dt>
    1646. 

  1646. <dd><p>Add a message to the log file</p>
    1647. 

  1647. </dd></dl>
    1648. 

  1648. 

  1649. <dl class="py function">
    1650. 

  1650. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.write_training_results">
    1651. 

  1651. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">write_training_results</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">writer</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">results_titles_list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">results_values_list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">epoch</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#write_training_results"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.write_training_results" title="Permalink to this definition"></a></dt>
    1652. 

  1652. <dd><p>Stores the training and validation loss and accuracy for current epoch in a tensorboard file</p>
    1653. 

  1653. </dd></dl>
    1654. 

  1654. 

  1655. <dl class="py function">
    1656. 

  1656. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.write_hpms">
    1657. 

  1657. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">write_hpms</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">writer</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">hpmstructs</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">[]</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">special_conf</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">{}</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#write_hpms"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.write_hpms" title="Permalink to this definition"></a></dt>
    1658. 

  1658. <dd><p>Stores the training and dataset hyper params in the tensorboard file</p>
    1659. 

  1659. </dd></dl>
    1660. 

  1660. 

  1661. <dl class="py function">
    1662. 

  1662. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.unpack_batch_items">
    1663. 

  1663. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">unpack_batch_items</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">batch_items</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#unpack_batch_items"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.unpack_batch_items" title="Permalink to this definition"></a></dt>
    1664. 

  1664. <dd><p>Adds support for unpacking batch items in train/validation loop.</p>
    1665. 

  1665. <dl>
    1666. 

  1666. <dt>&#64;param batch_items: (Union[tuple, torch.Tensor]) returned by the data loader, which is expected to be in one of</dt><dd><dl class="simple">
    1667. 

  1667. <dt>the following formats:</dt><dd><ol class="arabic simple">
    1668. 

  1668. <li><p>torch.Tensor or tuple, s.t inputs = batch_items[0], targets = batch_items[1] and len(batch_items) = 2</p></li>
    1669. 

  1669. <li><p>tuple: (inputs, targets, additional_batch_items)</p></li>
    1670. 

  1670. </ol>
    1671. 

  1671. </dd>
    1672. 

  1672. </dl>
    1673. 

  1673. <p>where inputs are fed to the network, targets are their corresponding labels and additional_batch_items is a
    1674. 

  1674. dictionary (format {additional_batch_item_i_name: additional_batch_item_i …}) which can be accessed through
    1675. 

  1675. the phase context under the attribute additional_batch_item_i_name, using a phase callback.</p>
    1676. 

  1676. </dd>
    1677. 

  1677. </dl>
    1678. 

  1678. <p>&#64;return: inputs, target, additional_batch_items</p>
    1679. 

  1679. </dd></dl>
    1680. 

  1680. 

  1681. <dl class="py function">
    1682. 

  1682. <dt class="sig sig-object py" id="super_gradients.training.utils.sg_model_utils.log_uncaught_exceptions">
    1683. 

  1683. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.sg_model_utils.</span></span><span class="sig-name descname"><span class="pre">log_uncaught_exceptions</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">logger</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/sg_model_utils.html#log_uncaught_exceptions"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.sg_model_utils.log_uncaught_exceptions" title="Permalink to this definition"></a></dt>
    1684. 

  1684. <dd><p>Makes logger log uncaught exceptions
    1685. 

  1685. &#64;param logger: logging.Logger</p>
    1686. 

  1686. <p>&#64;return: None</p>
    1687. 

  1687. </dd></dl>
    1688. 

  1688. 

  1689. </section>
    1690. 

  1690. <section id="module-super_gradients.training.utils.ssd_utils">
    1691. 

  1691. <span id="super-gradients-training-utils-ssd-utils-module"></span><h2>super_gradients.training.utils.ssd_utils module<a class="headerlink" href="#module-super_gradients.training.utils.ssd_utils" title="Permalink to this headline"></a></h2>
    1692. 

  1692. <dl class="py class">
    1693. 

  1693. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes">
    1694. 

  1694. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.ssd_utils.</span></span><span class="sig-name descname"><span class="pre">DefaultBoxes</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">fig_size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">feat_size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">steps</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">scales</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">aspect_ratios</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">scale_xy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">scale_wh</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.2</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#DefaultBoxes"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes" title="Permalink to this definition"></a></dt>
    1695. 

  1695. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1696. 

  1696. <p>Default Boxes, (aka: anchor boxes or priors boxes) used by SSD model</p>
    1697. 

  1697. <dl class="py property">
    1698. 

  1698. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes.scale_xy">
    1699. 

  1699. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">scale_xy</span></span><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes.scale_xy" title="Permalink to this definition"></a></dt>
    1700. 

  1700. <dd></dd></dl>
    1701. 

  1701. 

  1702. <dl class="py property">
    1703. 

  1703. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes.scale_wh">
    1704. 

  1704. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">scale_wh</span></span><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes.scale_wh" title="Permalink to this definition"></a></dt>
    1705. 

  1705. <dd></dd></dl>
    1706. 

  1706. 

  1707. <dl class="py method">
    1708. 

  1708. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes300_coco">
    1709. 

  1709. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">dboxes300_coco</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#DefaultBoxes.dboxes300_coco"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes300_coco" title="Permalink to this definition"></a></dt>
    1710. 

  1710. <dd></dd></dl>
    1711. 

  1711. 

  1712. <dl class="py method">
    1713. 

  1713. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes300_coco_from19">
    1714. 

  1714. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">dboxes300_coco_from19</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#DefaultBoxes.dboxes300_coco_from19"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes300_coco_from19" title="Permalink to this definition"></a></dt>
    1715. 

  1715. <dd><p>This dbox configuration is a bit different from the original dboxes300_coco
    1716. 

  1716. It is suitable for a network taking the first skip connection from a 19x19 layer (instead of 38x38 in the
    1717. 

  1717. original paper).
    1718. 

  1718. This offers less coverage for small objects but more aspect ratios options to larger objects (the original
    1719. 

  1719. paper supports object starting from size 21 pixels, while this config support objects starting from 60 pixels)</p>
    1720. 

  1720. </dd></dl>
    1721. 

  1721. 

  1722. <dl class="py method">
    1723. 

  1723. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes256_coco">
    1724. 

  1724. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">dboxes256_coco</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#DefaultBoxes.dboxes256_coco"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.DefaultBoxes.dboxes256_coco" title="Permalink to this definition"></a></dt>
    1725. 

  1725. <dd></dd></dl>
    1726. 

  1726. 

  1727. </dd></dl>
    1728. 

  1728. 

  1729. <dl class="py class">
    1730. 

  1730. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.SSDPostPredictCallback">
    1731. 

  1731. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.ssd_utils.</span></span><span class="sig-name descname"><span class="pre">SSDPostPredictCallback</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">conf:</span> <span class="pre">float</span> <span class="pre">=</span> <span class="pre">0.1</span></em>, <em class="sig-param"><span class="pre">iou:</span> <span class="pre">float</span> <span class="pre">=</span> <span class="pre">0.45</span></em>, <em class="sig-param"><span class="pre">classes:</span> <span class="pre">Optional[list]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">max_predictions:</span> <span class="pre">int</span> <span class="pre">=</span> <span class="pre">300</span></em>, <em class="sig-param"><span class="pre">nms_type:</span> <span class="pre">super_gradients.training.utils.detection_utils.NMS_Type</span> <span class="pre">=</span> <span class="pre">&lt;NMS_Type.ITERATIVE:</span> <span class="pre">'iterative'&gt;</span></em>, <em class="sig-param"><span class="pre">dboxes:</span> <span class="pre">super_gradients.training.utils.ssd_utils.DefaultBoxes</span> <span class="pre">=</span> <span class="pre">&lt;super_gradients.training.utils.ssd_utils.DefaultBoxes</span> <span class="pre">object&gt;</span></em>, <em class="sig-param"><span class="pre">device='cuda'</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#SSDPostPredictCallback"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.SSDPostPredictCallback" title="Permalink to this definition"></a></dt>
    1732. 

  1732. <dd><p>Bases: <a class="reference internal" href="#super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback" title="super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback</span></code></a></p>
    1733. 

  1733. <p>post prediction callback module to convert and filter predictions coming from the SSD net to a format
    1734. 

  1734. used by all other detection models</p>
    1735. 

  1735. <dl class="py method">
    1736. 

  1736. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.SSDPostPredictCallback.forward">
    1737. 

  1737. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/ssd_utils.html#SSDPostPredictCallback.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.SSDPostPredictCallback.forward" title="Permalink to this definition"></a></dt>
    1738. 

  1738. <dd><dl class="field-list simple">
    1739. 

  1739. <dt class="field-odd">Parameters</dt>
    1740. 

  1740. <dd class="field-odd"><ul class="simple">
    1741. 

  1741. <li><p><strong>x</strong> – the output of your model</p></li>
    1742. 

  1742. <li><p><strong>device</strong> – the device to move all output tensors into</p></li>
    1743. 

  1743. </ul>
    1744. 

  1744. </dd>
    1745. 

  1745. <dt class="field-even">Returns</dt>
    1746. 

  1746. <dd class="field-even"><p>a list with length batch_size, each item in the list is a detections
    1747. 

  1747. with shape: nx6 (x1, y1, x2, y2, confidence, class) where x and y are in range [0,1]</p>
    1748. 

  1748. </dd>
    1749. 

  1749. </dl>
    1750. 

  1750. </dd></dl>
    1751. 

  1751. 

  1752. <dl class="py attribute">
    1753. 

  1753. <dt class="sig sig-object py" id="super_gradients.training.utils.ssd_utils.SSDPostPredictCallback.training">
    1754. 

  1754. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.ssd_utils.SSDPostPredictCallback.training" title="Permalink to this definition"></a></dt>
    1755. 

  1755. <dd></dd></dl>
    1756. 

  1756. 

  1757. </dd></dl>
    1758. 

  1758. 

  1759. </section>
    1760. 

  1760. <section id="module-super_gradients.training.utils.utils">
    1761. 

  1761. <span id="super-gradients-training-utils-utils-module"></span><h2>super_gradients.training.utils.utils module<a class="headerlink" href="#module-super_gradients.training.utils.utils" title="Permalink to this headline"></a></h2>
    1762. 

  1762. <dl class="py function">
    1763. 

  1763. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.convert_to_tensor">
    1764. 

  1764. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">convert_to_tensor</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">array</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#convert_to_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.convert_to_tensor" title="Permalink to this definition"></a></dt>
    1765. 

  1765. <dd><p>Converts numpy arrays and lists to Torch tensors before calculation losses
    1766. 

  1766. :param array: torch.tensor / Numpy array / List</p>
    1767. 

  1767. </dd></dl>
    1768. 

  1768. 

  1769. <dl class="py class">
    1770. 

  1770. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.HpmStruct">
    1771. 

  1771. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">HpmStruct</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">entries</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.HpmStruct" title="Permalink to this definition"></a></dt>
    1772. 

  1772. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1773. 

  1773. <dl class="py method">
    1774. 

  1774. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.HpmStruct.set_schema">
    1775. 

  1775. <span class="sig-name descname"><span class="pre">set_schema</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">schema</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.set_schema"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.HpmStruct.set_schema" title="Permalink to this definition"></a></dt>
    1776. 

  1776. <dd></dd></dl>
    1777. 

  1777. 

  1778. <dl class="py method">
    1779. 

  1779. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.HpmStruct.override">
    1780. 

  1780. <span class="sig-name descname"><span class="pre">override</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">entries</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.override"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.HpmStruct.override" title="Permalink to this definition"></a></dt>
    1781. 

  1781. <dd></dd></dl>
    1782. 

  1782. 

  1783. <dl class="py method">
    1784. 

  1784. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.HpmStruct.to_dict">
    1785. 

  1785. <span class="sig-name descname"><span class="pre">to_dict</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.to_dict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.HpmStruct.to_dict" title="Permalink to this definition"></a></dt>
    1786. 

  1786. <dd></dd></dl>
    1787. 

  1787. 

  1788. <dl class="py method">
    1789. 

  1789. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.HpmStruct.validate">
    1790. 

  1790. <span class="sig-name descname"><span class="pre">validate</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.validate"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.HpmStruct.validate" title="Permalink to this definition"></a></dt>
    1791. 

  1791. <dd><p>Validate the current dict values according to the provided schema
    1792. 

  1792. :raises</p>
    1793. 

  1793. <blockquote>
    1794. 

  1794. <div><p><cite>AttributeError</cite> if schema was not set
    1795. 

  1795. <cite>jsonschema.exceptions.ValidationError</cite> if the instance is invalid
    1796. 

  1796. <cite>jsonschema.exceptions.SchemaError</cite> if the schema itselfis invalid</p>
    1797. 

  1797. </div></blockquote>
    1798. 

  1798. </dd></dl>
    1799. 

  1799. 

  1800. </dd></dl>
    1801. 

  1801. 

  1802. <dl class="py class">
    1803. 

  1803. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.WrappedModel">
    1804. 

  1804. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">WrappedModel</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">module</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#WrappedModel"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.WrappedModel" title="Permalink to this definition"></a></dt>
    1805. 

  1805. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1806. 

  1806. <dl class="py method">
    1807. 

  1807. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.WrappedModel.forward">
    1808. 

  1808. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#WrappedModel.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.WrappedModel.forward" title="Permalink to this definition"></a></dt>
    1809. 

  1809. <dd><p>Defines the computation performed at every call.</p>
    1810. 

  1810. <p>Should be overridden by all subclasses.</p>
    1811. 

  1811. <div class="admonition note">
    1812. 

  1812. <p class="admonition-title">Note</p>
    1813. 

  1813. <p>Although the recipe for forward pass needs to be defined within
    1814. 

  1814. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1815. 

  1815. instead of this since the former takes care of running the
    1816. 

  1816. registered hooks while the latter silently ignores them.</p>
    1817. 

  1817. </div>
    1818. 

  1818. </dd></dl>
    1819. 

  1819. 

  1820. <dl class="py attribute">
    1821. 

  1821. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.WrappedModel.training">
    1822. 

  1822. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.utils.WrappedModel.training" title="Permalink to this definition"></a></dt>
    1823. 

  1823. <dd></dd></dl>
    1824. 

  1824. 

  1825. </dd></dl>
    1826. 

  1826. 

  1827. <dl class="py class">
    1828. 

  1828. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.Timer">
    1829. 

  1829. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">Timer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.Timer" title="Permalink to this definition"></a></dt>
    1830. 

  1830. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1831. 

  1831. <p>A class to measure time handling both GPU &amp; CPU processes
    1832. 

  1832. Returns time in milliseconds</p>
    1833. 

  1833. <dl class="py method">
    1834. 

  1834. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.Timer.start">
    1835. 

  1835. <span class="sig-name descname"><span class="pre">start</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer.start"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.Timer.start" title="Permalink to this definition"></a></dt>
    1836. 

  1836. <dd></dd></dl>
    1837. 

  1837. 

  1838. <dl class="py method">
    1839. 

  1839. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.Timer.stop">
    1840. 

  1840. <span class="sig-name descname"><span class="pre">stop</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer.stop"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.Timer.stop" title="Permalink to this definition"></a></dt>
    1841. 

  1841. <dd></dd></dl>
    1842. 

  1842. 

  1843. </dd></dl>
    1844. 

  1844. 

  1845. <dl class="py class">
    1846. 

  1846. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.AverageMeter">
    1847. 

  1847. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">AverageMeter</span></span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#AverageMeter"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.AverageMeter" title="Permalink to this definition"></a></dt>
    1848. 

  1848. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1849. 

  1849. <p>A class to calculate the average of a metric, for each batch
    1850. 

  1850. during training/testing</p>
    1851. 

  1851. <dl class="py method">
    1852. 

  1852. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.AverageMeter.update">
    1853. 

  1853. <span class="sig-name descname"><span class="pre">update</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">value</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#AverageMeter.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.AverageMeter.update" title="Permalink to this definition"></a></dt>
    1854. 

  1854. <dd></dd></dl>
    1855. 

  1855. 

  1856. <dl class="py property">
    1857. 

  1857. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.AverageMeter.average">
    1858. 

  1858. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">average</span></span><a class="headerlink" href="#super_gradients.training.utils.utils.AverageMeter.average" title="Permalink to this definition"></a></dt>
    1859. 

  1859. <dd></dd></dl>
    1860. 

  1860. 

  1861. </dd></dl>
    1862. 

  1862. 

  1863. <dl class="py function">
    1864. 

  1864. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.tensor_container_to_device">
    1865. 

  1865. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">tensor_container_to_device</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">obj</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">dict</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">non_blocking</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#tensor_container_to_device"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.tensor_container_to_device" title="Permalink to this definition"></a></dt>
    1866. 

  1866. <dd><dl>
    1867. 

  1867. <dt>recursively send compounded objects to device (sending all tensors to device and maintaining structure)</dt><dd><p>:param obj           the object to send to device (list / tuple / tensor / dict)
    1868. 

  1868. :param device:       device to send the tensors to
    1869. 

  1869. :param non_blocking: used for DistributedDataParallel
    1870. 

  1870. :returns        an object with the same structure (tensors, lists, tuples) with the device pointers (like</p>
    1871. 

  1871. <blockquote>
    1872. 

  1872. <div><p>the return value of Tensor.to(device)</p>
    1873. 

  1873. </div></blockquote>
    1874. 

  1874. </dd>
    1875. 

  1875. </dl>
    1876. 

  1876. </dd></dl>
    1877. 

  1877. 

  1878. <dl class="py function">
    1879. 

  1879. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.get_param">
    1880. 

  1880. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">get_param</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">params</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">name</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">default_val</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#get_param"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.get_param" title="Permalink to this definition"></a></dt>
    1881. 

  1881. <dd><p>Retrieves a param from a parameter object/dict. If the parameter does not exist, will return default_val.
    1882. 

  1882. In case the default_val is of type dictionary, and a value is found in the params - the function
    1883. 

  1883. will return the default value dictionary with internal values overridden by the found value</p>
    1884. 

  1884. <p>i.e.
    1885. 

  1885. default_opt_params = {‘lr’:0.1, ‘momentum’:0.99, ‘alpha’:0.001}
    1886. 

  1886. training_params = {‘optimizer_params’: {‘lr’:0.0001}, ‘batch’: 32 …. }
    1887. 

  1887. get_param(training_params, name=’optimizer_params’, default_val=default_opt_params)
    1888. 

  1888. will return {‘lr’:0.0001, ‘momentum’:0.99, ‘alpha’:0.001}</p>
    1889. 

  1889. <dl class="field-list simple">
    1890. 

  1890. <dt class="field-odd">Parameters</dt>
    1891. 

  1891. <dd class="field-odd"><ul class="simple">
    1892. 

  1892. <li><p><strong>params</strong> – an object (typically HpmStruct) or a dict holding the params</p></li>
    1893. 

  1893. <li><p><strong>name</strong> – name of the searched parameter</p></li>
    1894. 

  1894. <li><p><strong>default_val</strong> – assumed to be the same type as the value searched in the params</p></li>
    1895. 

  1895. </ul>
    1896. 

  1896. </dd>
    1897. 

  1897. <dt class="field-even">Returns</dt>
    1898. 

  1898. <dd class="field-even"><p>the found value, or default if not found</p>
    1899. 

  1899. </dd>
    1900. 

  1900. </dl>
    1901. 

  1901. </dd></dl>
    1902. 

  1902. 

  1903. <dl class="py function">
    1904. 

  1904. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.static_vars">
    1905. 

  1905. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">static_vars</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">kwargs</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#static_vars"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.static_vars" title="Permalink to this definition"></a></dt>
    1906. 

  1906. <dd></dd></dl>
    1907. 

  1907. 

  1908. <dl class="py function">
    1909. 

  1909. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.print_once">
    1910. 

  1910. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">print_once</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">s</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#print_once"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.print_once" title="Permalink to this definition"></a></dt>
    1911. 

  1911. <dd></dd></dl>
    1912. 

  1912. 

  1913. <dl class="py function">
    1914. 

  1914. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.move_state_dict_to_device">
    1915. 

  1915. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">move_state_dict_to_device</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_sd</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#move_state_dict_to_device"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.move_state_dict_to_device" title="Permalink to this definition"></a></dt>
    1916. 

  1916. <dd><p>Moving model state dict tensors to target device (cuda or cpu)
    1917. 

  1917. :param model_sd: model state dict
    1918. 

  1918. :param device: either cuda or cpu</p>
    1919. 

  1919. </dd></dl>
    1920. 

  1920. 

  1921. <dl class="py function">
    1922. 

  1922. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.random_seed">
    1923. 

  1923. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">random_seed</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">is_ddp</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">seed</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#random_seed"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.random_seed" title="Permalink to this definition"></a></dt>
    1924. 

  1924. <dd><p>Sets random seed of numpy, torch and random.</p>
    1925. 

  1925. <p>When using ddp a seed will be set for each process according to its local rank derived from the device number.
    1926. 

  1926. :param is_ddp: bool, will set different random seed for each process when using ddp.
    1927. 

  1927. :param device: ‘cuda’,’cpu’, ‘cuda:&lt;device_number&gt;’
    1928. 

  1928. :param seed: int, random seed to be set</p>
    1929. 

  1929. </dd></dl>
    1930. 

  1930. 

  1931. <dl class="py function">
    1932. 

  1932. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.load_func">
    1933. 

  1933. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">load_func</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dotpath</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#load_func"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.load_func" title="Permalink to this definition"></a></dt>
    1934. 

  1934. <dd><p>load function in module.  function is right-most segment.</p>
    1935. 

  1935. <p>Used for passing functions (without calling them) in yaml files.</p>
    1936. 

  1936. <p>&#64;param dotpath: path to module.
    1937. 

  1937. &#64;return: a python function</p>
    1938. 

  1938. </dd></dl>
    1939. 

  1939. 

  1940. <dl class="py function">
    1941. 

  1941. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.get_filename_suffix_by_framework">
    1942. 

  1942. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">get_filename_suffix_by_framework</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">framework</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#get_filename_suffix_by_framework"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.get_filename_suffix_by_framework" title="Permalink to this definition"></a></dt>
    1943. 

  1943. <dd><p>Return the file extension of framework.</p>
    1944. 

  1944. <p>&#64;param framework: (str)
    1945. 

  1945. &#64;return: (str) the suffix for the specific framework</p>
    1946. 

  1946. </dd></dl>
    1947. 

  1947. 

  1948. <dl class="py function">
    1949. 

  1949. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.check_models_have_same_weights">
    1950. 

  1950. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">check_models_have_same_weights</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_1</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_2</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.module.Module</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#check_models_have_same_weights"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.check_models_have_same_weights" title="Permalink to this definition"></a></dt>
    1951. 

  1951. <dd><p>Checks whether two networks have the same weights</p>
    1952. 

  1952. <p>&#64;param model_1: Net to be checked
    1953. 

  1953. &#64;param model_2: Net to be checked
    1954. 

  1954. &#64;return: True iff the two networks have the same weights</p>
    1955. 

  1955. </dd></dl>
    1956. 

  1956. 

  1957. <dl class="py function">
    1958. 

  1958. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.recursive_override">
    1959. 

  1959. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">recursive_override</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">base</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">extension</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#recursive_override"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.recursive_override" title="Permalink to this definition"></a></dt>
    1960. 

  1960. <dd></dd></dl>
    1961. 

  1961. 

  1962. <dl class="py function">
    1963. 

  1963. <dt class="sig sig-object py" id="super_gradients.training.utils.utils.download_and_unzip_from_url">
    1964. 

  1964. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.utils.</span></span><span class="sig-name descname"><span class="pre">download_and_unzip_from_url</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">url</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dir</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'.'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">unzip</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">delete</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#download_and_unzip_from_url"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.utils.download_and_unzip_from_url" title="Permalink to this definition"></a></dt>
    1965. 

  1965. <dd><p>Downloads a zip file from url to dir, and unzips it.</p>
    1966. 

  1966. <dl class="field-list simple">
    1967. 

  1967. <dt class="field-odd">Parameters</dt>
    1968. 

  1968. <dd class="field-odd"><ul class="simple">
    1969. 

  1969. <li><p><strong>url</strong> – Url to download the file from.</p></li>
    1970. 

  1970. <li><p><strong>dir</strong> – Destination directory.</p></li>
    1971. 

  1971. <li><p><strong>unzip</strong> – Whether to unzip the downloaded file.</p></li>
    1972. 

  1972. <li><p><strong>delete</strong> – Whether to delete the zip file.</p></li>
    1973. 

  1973. </ul>
    1974. 

  1974. </dd>
    1975. 

  1975. </dl>
    1976. 

  1976. <p>used to downlaod VOC.</p>
    1977. 

  1977. <p>Source:
    1978. 

  1978. <a class="reference external" href="https://github.com/ultralytics/yolov5/blob/master/data/VOC.yaml">https://github.com/ultralytics/yolov5/blob/master/data/VOC.yaml</a></p>
    1979. 

  1979. </dd></dl>
    1980. 

  1980. 

  1981. </section>
    1982. 

  1982. <section id="module-super_gradients.training.utils.weight_averaging_utils">
    1983. 

  1983. <span id="super-gradients-training-utils-weight-averaging-utils-module"></span><h2>super_gradients.training.utils.weight_averaging_utils module<a class="headerlink" href="#module-super_gradients.training.utils.weight_averaging_utils" title="Permalink to this headline"></a></h2>
    1984. 

  1984. <dl class="py class">
    1985. 

  1985. <dt class="sig sig-object py" id="super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging">
    1986. 

  1986. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.weight_averaging_utils.</span></span><span class="sig-name descname"><span class="pre">ModelWeightAveraging</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ckpt_dir</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">greater_is_better</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">source_ckpt_folder_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metric_to_watch</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'acc'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metric_idx</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">load_checkpoint</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">number_of_models_to_average</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">model_checkpoints_location</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'local'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/weight_averaging_utils.html#ModelWeightAveraging"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging" title="Permalink to this definition"></a></dt>
    1987. 

  1987. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1988. 

  1988. <p>Utils class for managing the averaging of the best several snapshots into a single model.
    1989. 

  1989. A snapshot dictionary file and the average model will be saved / updated at every epoch and evaluated only when
    1990. 

  1990. training is completed. The snapshot file will only be deleted upon completing the training.
    1991. 

  1991. The snapshot dict will be managed on cpu.</p>
    1992. 

  1992. <dl class="py method">
    1993. 

  1993. <dt class="sig sig-object py" id="super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.update_snapshots_dict">
    1994. 

  1994. <span class="sig-name descname"><span class="pre">update_snapshots_dict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">validation_results_tuple</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/weight_averaging_utils.html#ModelWeightAveraging.update_snapshots_dict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.update_snapshots_dict" title="Permalink to this definition"></a></dt>
    1995. 

  1995. <dd><p>Update the snapshot dict and returns the updated average model for saving
    1996. 

  1996. :param model: the latest model
    1997. 

  1997. :param validation_results_tuple: performance of the latest model</p>
    1998. 

  1998. </dd></dl>
    1999. 

  1999. 

  2000. <dl class="py method">
    2001. 

  2001. <dt class="sig sig-object py" id="super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.get_average_model">
    2002. 

  2002. <span class="sig-name descname"><span class="pre">get_average_model</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">validation_results_tuple</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/weight_averaging_utils.html#ModelWeightAveraging.get_average_model"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.get_average_model" title="Permalink to this definition"></a></dt>
    2003. 

  2003. <dd><p>Returns the averaged model
    2004. 

  2004. :param model: will be used to determine arch
    2005. 

  2005. :param validation_results_tuple: if provided, will update the average model before returning
    2006. 

  2006. :param target_device: if provided, return sd on target device</p>
    2007. 

  2007. </dd></dl>
    2008. 

  2008. 

  2009. <dl class="py method">
    2010. 

  2010. <dt class="sig sig-object py" id="super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.cleanup">
    2011. 

  2011. <span class="sig-name descname"><span class="pre">cleanup</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/weight_averaging_utils.html#ModelWeightAveraging.cleanup"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.weight_averaging_utils.ModelWeightAveraging.cleanup" title="Permalink to this definition"></a></dt>
    2012. 

  2012. <dd><p>Delete snapshot file when reaching the last epoch</p>
    2013. 

  2013. </dd></dl>
    2014. 

  2014. 

  2015. </dd></dl>
    2016. 

  2016. 

  2017. </section>
    2018. 

  2018. <section id="module-super_gradients.training.utils">
    2019. 

  2019. <span id="module-contents"></span><h2>Module contents<a class="headerlink" href="#module-super_gradients.training.utils" title="Permalink to this headline"></a></h2>
    2020. 

  2020. <dl class="py class">
    2021. 

  2021. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer">
    2022. 

  2022. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">Timer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.Timer" title="Permalink to this definition"></a></dt>
    2023. 

  2023. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    2024. 

  2024. <p>A class to measure time handling both GPU &amp; CPU processes
    2025. 

  2025. Returns time in milliseconds</p>
    2026. 

  2026. <dl class="py method">
    2027. 

  2027. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer.start">
    2028. 

  2028. <span class="sig-name descname"><span class="pre">start</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer.start"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.Timer.start" title="Permalink to this definition"></a></dt>
    2029. 

  2029. <dd></dd></dl>
    2030. 

  2030. 

  2031. <dl class="py method">
    2032. 

  2032. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer.stop">
    2033. 

  2033. <span class="sig-name descname"><span class="pre">stop</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#Timer.stop"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.Timer.stop" title="Permalink to this definition"></a></dt>
    2034. 

  2034. <dd></dd></dl>
    2035. 

  2035. 

  2036. </dd></dl>
    2037. 

  2037. 

  2038. <dl class="py class">
    2039. 

  2039. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct">
    2040. 

  2040. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">HpmStruct</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">entries</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.HpmStruct" title="Permalink to this definition"></a></dt>
    2041. 

  2041. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    2042. 

  2042. <dl class="py method">
    2043. 

  2043. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.set_schema">
    2044. 

  2044. <span class="sig-name descname"><span class="pre">set_schema</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">schema</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.set_schema"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.HpmStruct.set_schema" title="Permalink to this definition"></a></dt>
    2045. 

  2045. <dd></dd></dl>
    2046. 

  2046. 

  2047. <dl class="py method">
    2048. 

  2048. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.override">
    2049. 

  2049. <span class="sig-name descname"><span class="pre">override</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">**</span></span><span class="n"><span class="pre">entries</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.override"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.HpmStruct.override" title="Permalink to this definition"></a></dt>
    2050. 

  2050. <dd></dd></dl>
    2051. 

  2051. 

  2052. <dl class="py method">
    2053. 

  2053. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.to_dict">
    2054. 

  2054. <span class="sig-name descname"><span class="pre">to_dict</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.to_dict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.HpmStruct.to_dict" title="Permalink to this definition"></a></dt>
    2055. 

  2055. <dd></dd></dl>
    2056. 

  2056. 

  2057. <dl class="py method">
    2058. 

  2058. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.validate">
    2059. 

  2059. <span class="sig-name descname"><span class="pre">validate</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#HpmStruct.validate"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.HpmStruct.validate" title="Permalink to this definition"></a></dt>
    2060. 

  2060. <dd><p>Validate the current dict values according to the provided schema
    2061. 

  2061. :raises</p>
    2062. 

  2062. <blockquote>
    2063. 

  2063. <div><p><cite>AttributeError</cite> if schema was not set
    2064. 

  2064. <cite>jsonschema.exceptions.ValidationError</cite> if the instance is invalid
    2065. 

  2065. <cite>jsonschema.exceptions.SchemaError</cite> if the schema itselfis invalid</p>
    2066. 

  2066. </div></blockquote>
    2067. 

  2067. </dd></dl>
    2068. 

  2068. 

  2069. </dd></dl>
    2070. 

  2070. 

  2071. <dl class="py class">
    2072. 

  2072. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel">
    2073. 

  2073. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">WrappedModel</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">module</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#WrappedModel"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.WrappedModel" title="Permalink to this definition"></a></dt>
    2074. 

  2074. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    2075. 

  2075. <dl class="py method">
    2076. 

  2076. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel.forward">
    2077. 

  2077. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">x</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#WrappedModel.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.WrappedModel.forward" title="Permalink to this definition"></a></dt>
    2078. 

  2078. <dd><p>Defines the computation performed at every call.</p>
    2079. 

  2079. <p>Should be overridden by all subclasses.</p>
    2080. 

  2080. <div class="admonition note">
    2081. 

  2081. <p class="admonition-title">Note</p>
    2082. 

  2082. <p>Although the recipe for forward pass needs to be defined within
    2083. 

  2083. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    2084. 

  2084. instead of this since the former takes care of running the
    2085. 

  2085. registered hooks while the latter silently ignores them.</p>
    2086. 

  2086. </div>
    2087. 

  2087. </dd></dl>
    2088. 

  2088. 

  2089. <dl class="py attribute">
    2090. 

  2090. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel.training">
    2091. 

  2091. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.utils.WrappedModel.training" title="Permalink to this definition"></a></dt>
    2092. 

  2092. <dd></dd></dl>
    2093. 

  2093. 

  2094. </dd></dl>
    2095. 

  2095. 

  2096. <dl class="py function">
    2097. 

  2097. <dt class="sig sig-object py" id="super_gradients.training.utils.convert_to_tensor">
    2098. 

  2098. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">convert_to_tensor</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">array</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#convert_to_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.convert_to_tensor" title="Permalink to this definition"></a></dt>
    2099. 

  2099. <dd><p>Converts numpy arrays and lists to Torch tensors before calculation losses
    2100. 

  2100. :param array: torch.tensor / Numpy array / List</p>
    2101. 

  2101. </dd></dl>
    2102. 

  2102. 

  2103. <dl class="py function">
    2104. 

  2104. <dt class="sig sig-object py" id="super_gradients.training.utils.get_param">
    2105. 

  2105. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">get_param</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">params</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">name</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">default_val</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#get_param"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.get_param" title="Permalink to this definition"></a></dt>
    2106. 

  2106. <dd><p>Retrieves a param from a parameter object/dict. If the parameter does not exist, will return default_val.
    2107. 

  2107. In case the default_val is of type dictionary, and a value is found in the params - the function
    2108. 

  2108. will return the default value dictionary with internal values overridden by the found value</p>
    2109. 

  2109. <p>i.e.
    2110. 

  2110. default_opt_params = {‘lr’:0.1, ‘momentum’:0.99, ‘alpha’:0.001}
    2111. 

  2111. training_params = {‘optimizer_params’: {‘lr’:0.0001}, ‘batch’: 32 …. }
    2112. 

  2112. get_param(training_params, name=’optimizer_params’, default_val=default_opt_params)
    2113. 

  2113. will return {‘lr’:0.0001, ‘momentum’:0.99, ‘alpha’:0.001}</p>
    2114. 

  2114. <dl class="field-list simple">
    2115. 

  2115. <dt class="field-odd">Parameters</dt>
    2116. 

  2116. <dd class="field-odd"><ul class="simple">
    2117. 

  2117. <li><p><strong>params</strong> – an object (typically HpmStruct) or a dict holding the params</p></li>
    2118. 

  2118. <li><p><strong>name</strong> – name of the searched parameter</p></li>
    2119. 

  2119. <li><p><strong>default_val</strong> – assumed to be the same type as the value searched in the params</p></li>
    2120. 

  2120. </ul>
    2121. 

  2121. </dd>
    2122. 

  2122. <dt class="field-even">Returns</dt>
    2123. 

  2123. <dd class="field-even"><p>the found value, or default if not found</p>
    2124. 

  2124. </dd>
    2125. 

  2125. </dl>
    2126. 

  2126. </dd></dl>
    2127. 

  2127. 

  2128. <dl class="py function">
    2129. 

  2129. <dt class="sig sig-object py" id="super_gradients.training.utils.tensor_container_to_device">
    2130. 

  2130. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">tensor_container_to_device</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">obj</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">tuple</span><span class="p"><span class="pre">,</span> </span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">dict</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">non_blocking</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#tensor_container_to_device"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.tensor_container_to_device" title="Permalink to this definition"></a></dt>
    2131. 

  2131. <dd><dl>
    2132. 

  2132. <dt>recursively send compounded objects to device (sending all tensors to device and maintaining structure)</dt><dd><p>:param obj           the object to send to device (list / tuple / tensor / dict)
    2133. 

  2133. :param device:       device to send the tensors to
    2134. 

  2134. :param non_blocking: used for DistributedDataParallel
    2135. 

  2135. :returns        an object with the same structure (tensors, lists, tuples) with the device pointers (like</p>
    2136. 

  2136. <blockquote>
    2137. 

  2137. <div><p>the return value of Tensor.to(device)</p>
    2138. 

  2138. </div></blockquote>
    2139. 

  2139. </dd>
    2140. 

  2140. </dl>
    2141. 

  2141. </dd></dl>
    2142. 

  2142. 

  2143. <dl class="py function">
    2144. 

  2144. <dt class="sig sig-object py" id="super_gradients.training.utils.adapt_state_dict_to_fit_model_layer_names">
    2145. 

  2145. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">adapt_state_dict_to_fit_model_layer_names</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_state_dict</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">source_ckpt</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exclude</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">[]</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#adapt_state_dict_to_fit_model_layer_names"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.adapt_state_dict_to_fit_model_layer_names" title="Permalink to this definition"></a></dt>
    2146. 

  2146. <dd><p>Given a model state dict and source checkpoints, the method tries to correct the keys in the model_state_dict to fit
    2147. 

  2147. the ckpt in order to properly load the weights into the model. If unsuccessful - returns None</p>
    2148. 

  2148. <blockquote>
    2149. 

  2149. <div><dl class="field-list simple">
    2150. 

  2150. <dt class="field-odd">param model_state_dict</dt>
    2151. 

  2151. <dd class="field-odd"><p>the model state_dict</p>
    2152. 

  2152. </dd>
    2153. 

  2153. <dt class="field-even">param source_ckpt</dt>
    2154. 

  2154. <dd class="field-even"><p>checkpoint dict</p>
    2155. 

  2155. </dd>
    2156. 

  2156. </dl>
    2157. 

  2157. <p>:exclude                  optional list for excluded layers
    2158. 

  2158. :return: renamed checkpoint dict (if possible)</p>
    2159. 

  2159. </div></blockquote>
    2160. 

  2160. </dd></dl>
    2161. 

  2161. 

  2162. <dl class="py function">
    2163. 

  2163. <dt class="sig sig-object py" id="super_gradients.training.utils.raise_informative_runtime_error">
    2164. 

  2164. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">raise_informative_runtime_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">state_dict</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">checkpoint</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">exception_msg</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/checkpoint_utils.html#raise_informative_runtime_error"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.raise_informative_runtime_error" title="Permalink to this definition"></a></dt>
    2165. 

  2165. <dd><p>Given a model state dict and source checkpoints, the method calls “adapt_state_dict_to_fit_model_layer_names”
    2166. 

  2166. and enhances the exception_msg if loading the checkpoint_dict via the conversion method is possible</p>
    2167. 

  2167. </dd></dl>
    2168. 

  2168. 

  2169. <dl class="py function">
    2170. 

  2170. <dt class="sig sig-object py" id="super_gradients.training.utils.random_seed">
    2171. 

  2171. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.utils.</span></span><span class="sig-name descname"><span class="pre">random_seed</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">is_ddp</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">seed</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/utils/utils.html#random_seed"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.utils.random_seed" title="Permalink to this definition"></a></dt>
    2172. 

  2172. <dd><p>Sets random seed of numpy, torch and random.</p>
    2173. 

  2173. <p>When using ddp a seed will be set for each process according to its local rank derived from the device number.
    2174. 

  2174. :param is_ddp: bool, will set different random seed for each process when using ddp.
    2175. 

  2175. :param device: ‘cuda’,’cpu’, ‘cuda:&lt;device_number&gt;’
    2176. 

  2176. :param seed: int, random seed to be set</p>
    2177. 

  2177. </dd></dl>
    2178. 

  2178. 

  2179. </section>
    2180. 

  2180. </section>
    2181. 

  2181. 

  2182. 

  2183.            </div>
    2184. 

  2184.           </div>
    2185. 

  2185.           <footer>
    2186. 

  2186. 

  2187.   <hr/>
    2188. 

  2188. 

  2189.   <div role="contentinfo">
    2190. 

  2190.     <p>&#169; Copyright 2021, SuperGradients team.</p>
    2191. 

  2191.   </div>
    2192. 

  2192. 

  2193.   Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    2194. 

  2194.     <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    2195. 

  2195.     provided by <a href="https://readthedocs.org">Read the Docs</a>.
    2196. 

  2196.    
    2197. 

  2197. 

  2198. </footer>
    2199. 

  2199.         </div>
    2200. 

  2200.       </div>
    2201. 

  2201.     </section>
    2202. 

  2202.   </div>
    2203. 

  2203.   <script>
    2204. 

  2204.       jQuery(function () {
    2205. 

  2205.           SphinxRtdTheme.Navigation.enable(true);
    2206. 

  2206.       });
    2207. 

  2207.   </script> 
    2208. 

  2208. 

  2209. </body>
    2210. 

  2210. </html>
    2211.
Prev
Next
Tip!

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

Comments

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

Use AWS S3 as storage!

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

Specify your S3 bucket

Access key (If needed)

Congratulations!

super-gradients is now integrated with AWS S3!

Delete Storage Key

Are you sure you want to delete this access key?

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

Use Google Cloud Storage!

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

Specify your Google Storage bucket

Service Account Key

Congratulations!

super-gradients is now integrated with Google Cloud Storage!

Delete Storage Key

Are you sure you want to delete this access key?

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

Use Azure Cloud Storage!

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

Specify your Azure Storage bucket

Access key (If needed)

Congratulations!

super-gradients is now integrated with Azure Cloud Storage!

Delete Storage Key

Are you sure you want to delete this access key?

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

Use any S3 compatible storage!

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

Specify your S3 bucket

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

Access key (If needed)

Congratulations!

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

Delete Storage Key

Are you sure you want to delete this access key?

No
Yes