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

#367 fix: Request correct hydra-core

Merged
Ghost merged 1 commits into Deci-AI:master from deci-ai:hotfix/ALG-000_hydra-req
Conversation 0 Files changed 1 Commits 1
File Compare
super-gradients
/
docs
/
super_gradients.training.html
View left file View right file
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
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
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055 
  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>Training 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.     <link rel="next" title="What is SuperGradients?" href="user_guide.html" />
    23. 

  23.     <link rel="prev" title="Common package" href="super_gradients.common.html" /> 
    24. 

  24. </head>
    25. 

  25. 

  26. <body class="wy-body-for-nav"> 
    27. 

  27.   <div class="wy-grid-for-nav">
    28. 

  28.     <nav data-toggle="wy-nav-shift" class="wy-nav-side">
    29. 

  29.       <div class="wy-side-scroll">
    30. 

  30.         <div class="wy-side-nav-search" >
    31. 

  31.             <a href="index.html" class="icon icon-home"> SuperGradients
    32. 

  32.           </a>
    33. 

  33. <div role="search">
    34. 

  34.   <form id="rtd-search-form" class="wy-form" action="search.html" method="get">
    35. 

  35.     <input type="text" name="q" placeholder="Search docs" />
    36. 

  36.     <input type="hidden" name="check_keywords" value="yes" />
    37. 

  37.     <input type="hidden" name="area" value="default" />
    38. 

  38.   </form>
    39. 

  39. </div>
    40. 

  40.         </div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
    41. 

  41.               <p class="caption"><span class="caption-text">Welcome To SuperGradients</span></p>
    42. 

  42. <ul>
    43. 

  43. <li class="toctree-l1"><a class="reference internal" href="welcome.html">Fill our 4-question quick survey! We will raffle free SuperGradients swag between those who will participate -&gt; Fill Survey</a></li>
    44. 

  44. <li class="toctree-l1"><a class="reference internal" href="welcome.html#supergradients">SuperGradients</a></li>
    45. 

  45. </ul>
    46. 

  46. <p class="caption"><span class="caption-text">Technical Documentation</span></p>
    47. 

  47. <ul class="current">
    48. 

  48. <li class="toctree-l1"><a class="reference internal" href="super_gradients.common.html">Common package</a></li>
    49. 

  49. <li class="toctree-l1 current"><a class="current reference internal" href="#">Training package</a><ul>
    50. 

  50. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training">super_gradients.training module</a></li>
    51. 

  51. <li class="toctree-l2"><a class="reference internal" href="#super-gradients-training-datasets-module">super_gradients.training.datasets module</a></li>
    52. 

  52. <li class="toctree-l2"><a class="reference internal" href="#super-gradients-training-exceptions-module">super_gradients.training.exceptions module</a></li>
    53. 

  53. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.legacy">super_gradients.training.legacy module</a></li>
    54. 

  54. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.losses">super_gradients.training.losses_models module</a></li>
    55. 

  55. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.metrics">super_gradients.training.metrics module</a></li>
    56. 

  56. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.models">super_gradients.training.models module</a></li>
    57. 

  57. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.sg_model">super_gradients.training.sg_model module</a></li>
    58. 

  58. <li class="toctree-l2"><a class="reference internal" href="#module-super_gradients.training.utils">super_gradients.training.utils module</a></li>
    59. 

  59. <li class="toctree-l2"><a class="reference internal" href="#module-contents">Module contents</a></li>
    60. 

  60. </ul>
    61. 

  61. </li>
    62. 

  62. </ul>
    63. 

  63. <p class="caption"><span class="caption-text">User Guide</span></p>
    64. 

  64. <ul>
    65. 

  65. <li class="toctree-l1"><a class="reference internal" href="user_guide.html">What is SuperGradients?</a></li>
    66. 

  66. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#introducing-the-supergradients-library">Introducing the SuperGradients library</a></li>
    67. 

  67. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#installation">Installation</a></li>
    68. 

  68. <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>
    69. 

  69. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#training-parameters">Training Parameters</a></li>
    70. 

  70. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#logs-and-checkpoints">Logs and Checkpoints</a></li>
    71. 

  71. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#dataset-parameters">Dataset Parameters</a></li>
    72. 

  72. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#network-architectures">Network Architectures</a></li>
    73. 

  73. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#pretrained-models">Pretrained Models</a></li>
    74. 

  74. <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>
    75. 

  75. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#professional-tools-integration">Professional Tools Integration</a></li>
    76. 

  76. <li class="toctree-l1"><a class="reference internal" href="user_guide.html#supergradients-faq">SuperGradients FAQ</a></li>
    77. 

  77. </ul>
    78. 

  78. 

  79.         </div>
    80. 

  80.       </div>
    81. 

  81.     </nav>
    82. 

  82. 

  83.     <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
    84. 

  84.           <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
    85. 

  85.           <a href="index.html">SuperGradients</a>
    86. 

  86.       </nav>
    87. 

  87. 

  88.       <div class="wy-nav-content">
    89. 

  89.         <div class="rst-content">
    90. 

  90.           <div role="navigation" aria-label="Page navigation">
    91. 

  91.   <ul class="wy-breadcrumbs">
    92. 

  92.       <li><a href="index.html" class="icon icon-home"></a> &raquo;</li>
    93. 

  93.       <li>Training package</li>
    94. 

  94.       <li class="wy-breadcrumbs-aside">
    95. 

  95.             <a href="_sources/super_gradients.training.rst.txt" rel="nofollow"> View page source</a>
    96. 

  96.       </li>
    97. 

  97.   </ul>
    98. 

  98.   <hr/>
    99. 

  99. </div>
    100. 

  100.           <div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
    101. 

  101.            <div itemprop="articleBody">
    102. 

  102.              
    103. 

  103.   <section id="training-package">
    104. 

  104. <h1>Training package<a class="headerlink" href="#training-package" title="Permalink to this headline"></a></h1>
    105. 

  105. <table class="longtable docutils align-default">
    106. 

  106. <colgroup>
    107. 

  107. <col style="width: 10%" />
    108. 

  108. <col style="width: 90%" />
    109. 

  109. </colgroup>
    110. 

  110. <tbody>
    111. 

  111. </tbody>
    112. 

  112. </table>
    113. 

  113. <section id="module-super_gradients.training">
    114. 

  114. <span id="super-gradients-training-module"></span><h2>super_gradients.training module<a class="headerlink" href="#module-super_gradients.training" title="Permalink to this headline"></a></h2>
    115. 

  115. <dl class="py class">
    116. 

  116. <dt class="sig sig-object py" id="super_gradients.training.DataAugmentation">
    117. 

  117. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">DataAugmentation</span></span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.DataAugmentation" title="Permalink to this definition"></a></dt>
    118. 

  118. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    119. 

  119. <dl class="py method">
    120. 

  120. <dt class="sig sig-object py" id="super_gradients.training.DataAugmentation.to_tensor">
    121. 

  121. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">to_tensor</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.to_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.DataAugmentation.to_tensor" title="Permalink to this definition"></a></dt>
    122. 

  122. <dd></dd></dl>
    123. 

  123. 

  124. <dl class="py method">
    125. 

  125. <dt class="sig sig-object py" id="super_gradients.training.DataAugmentation.normalize">
    126. 

  126. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">normalize</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">mean</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">std</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.normalize"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.DataAugmentation.normalize" title="Permalink to this definition"></a></dt>
    127. 

  127. <dd></dd></dl>
    128. 

  128. 

  129. <dl class="py method">
    130. 

  130. <dt class="sig sig-object py" id="super_gradients.training.DataAugmentation.cutout">
    131. 

  131. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">cutout</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">mask_size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p</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">cutout_inside</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">mask_color</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">(0,</span> <span class="pre">0,</span> <span class="pre">0)</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.cutout"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.DataAugmentation.cutout" title="Permalink to this definition"></a></dt>
    132. 

  132. <dd></dd></dl>
    133. 

  133. 

  134. </dd></dl>
    135. 

  135. 

  136. <dl class="py class">
    137. 

  137. <dt class="sig sig-object py" id="super_gradients.training.TestDatasetInterface">
    138. 

  138. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">TestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">trainset</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dataset_params</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">classes</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/datasets/dataset_interfaces/dataset_interface.html#TestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.TestDatasetInterface" title="Permalink to this definition"></a></dt>
    139. 

  139. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    140. 

  140. <dl class="py method">
    141. 

  141. <dt class="sig sig-object py" id="super_gradients.training.TestDatasetInterface.get_data_loaders">
    142. 

  142. <span class="sig-name descname"><span class="pre">get_data_loaders</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">batch_size_factor</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">num_workers</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">8</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">train_batch_size</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">val_batch_size</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">distributed_sampler</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/datasets/dataset_interfaces/dataset_interface.html#TestDatasetInterface.get_data_loaders"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.TestDatasetInterface.get_data_loaders" title="Permalink to this definition"></a></dt>
    143. 

  143. <dd><p>Get self.train_loader, self.val_loader, self.test_loader, self.classes.</p>
    144. 

  144. <p>If the data loaders haven’t been initialized yet, build them first.</p>
    145. 

  145. <dl class="field-list simple">
    146. 

  146. <dt class="field-odd">Parameters</dt>
    147. 

  147. <dd class="field-odd"><p><strong>kwargs</strong> – kwargs are passed to build_data_loaders.</p>
    148. 

  148. </dd>
    149. 

  149. </dl>
    150. 

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

  151. 

  152. </dd></dl>
    153. 

  153. 

  154. <dl class="py class">
    155. 

  155. <dt class="sig sig-object py" id="super_gradients.training.SgModel">
    156. 

  156. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">SgModel</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">experiment_name:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">device:</span> <span class="pre">Optional[str]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">multi_gpu:</span> <span class="pre">Union[super_gradients.common.data_types.enum.multi_gpu_mode.MultiGPUMode</span></em>, <em class="sig-param"><span class="pre">str]</span> <span class="pre">=</span> <span class="pre">&lt;MultiGPUMode.OFF:</span> <span class="pre">'Off'&gt;</span></em>, <em class="sig-param"><span class="pre">model_checkpoints_location:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'local'</span></em>, <em class="sig-param"><span class="pre">overwrite_local_checkpoint:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></em>, <em class="sig-param"><span class="pre">ckpt_name:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'ckpt_latest.pth'</span></em>, <em class="sig-param"><span class="pre">post_prediction_callback:</span> <span class="pre">Optional[super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">ckpt_root_dir:</span> <span class="pre">Optional[str]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">train_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">valid_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">test_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">classes:</span> <span class="pre">Optional[List[Any]]</span> <span class="pre">=</span> <span class="pre">None</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel" title="Permalink to this definition"></a></dt>
    157. 

  157. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    158. 

  158. <p>SuperGradient Model - Base Class for Sg Models</p>
    159. 

  159. <dl class="py method">
    160. 

  160. <dt class="sig sig-object py" id="super_gradients.training.SgModel.train">
    161. 

  161. <span class="sig-name descname"><span class="pre">train</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_epochs</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">initial_epoch</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">save_model</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/sg_model/sg_model.html#SgModel.train"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.train" title="Permalink to this definition"></a></dt>
    162. 

  162. <dd><p>the main function used for the training, h.p. updating, logging etc.</p>
    163. 

  163. </dd></dl>
    164. 

  164. 

  165. <dl class="py method">
    166. 

  166. <dt class="sig sig-object py" id="super_gradients.training.SgModel.predict">
    167. 

  167. <span class="sig-name descname"><span class="pre">predict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">idx</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/sg_model/sg_model.html#SgModel.predict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.predict" title="Permalink to this definition"></a></dt>
    168. 

  168. <dd><p>returns the predictions and label of the current inputs</p>
    169. 

  169. </dd></dl>
    170. 

  170. 

  171. <dl class="py method">
    172. 

  172. <dt class="sig sig-object py">
    173. 

  173. <span class="sig-name descname"><span class="pre">test(epoch</span> <span class="pre">:</span> <span class="pre">int,</span> <span class="pre">idx</span> <span class="pre">:</span> <span class="pre">int,</span> <span class="pre">save</span> <span class="pre">:</span> <span class="pre">bool):</span></span></dt>
    174. 

  174. <dd><p>returns the test loss, accuracy and runtime</p>
    175. 

  175. </dd></dl>
    176. 

  176. 

  177. <dl class="py method">
    178. 

  178. <dt class="sig sig-object py" id="super_gradients.training.SgModel.connect_dataset_interface">
    179. 

  179. <span class="sig-name descname"><span class="pre">connect_dataset_interface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_interface</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">data_loader_num_workers</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">8</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.connect_dataset_interface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.connect_dataset_interface" title="Permalink to this definition"></a></dt>
    180. 

  180. <dd><dl class="field-list simple">
    181. 

  181. <dt class="field-odd">Parameters</dt>
    182. 

  182. <dd class="field-odd"><ul class="simple">
    183. 

  183. <li><p><strong>dataset_interface</strong> – DatasetInterface object</p></li>
    184. 

  184. <li><p><strong>data_loader_num_workers</strong> – The number of threads to initialize the Data Loaders with
    185. 

  185. The dataset to be connected</p></li>
    186. 

  186. </ul>
    187. 

  187. </dd>
    188. 

  188. </dl>
    189. 

  189. </dd></dl>
    190. 

  190. 

  191. <dl class="py method">
    192. 

  192. <dt class="sig sig-object py" id="super_gradients.training.SgModel.build_model">
    193. 

  193. <span class="sig-name descname"><span class="pre">build_model</span></span><span class="sig-paren">(</span><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">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.nn.modules.module.Module</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">arch_params</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">checkpoint_params</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="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</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/sg_model/sg_model.html#SgModel.build_model"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.build_model" title="Permalink to this definition"></a></dt>
    194. 

  194. <dd><dl class="field-list simple">
    195. 

  195. <dt class="field-odd">Parameters</dt>
    196. 

  196. <dd class="field-odd"><ul class="simple">
    197. 

  197. <li><p><strong>architecture</strong> – Defines the network’s architecture from models/ALL_ARCHITECTURES</p></li>
    198. 

  198. <li><p><strong>arch_params</strong> – Architecture H.P. e.g.: block, num_blocks, num_classes, etc.</p></li>
    199. 

  199. <li><p><strong>checkpoint_params</strong> – <p>Dictionary like object with the following key:values:</p>
    200. 

  200. <p>load_checkpoint:            Load a pre-trained checkpoint
    201. 

  201. strict_load:                See StrictLoad class documentation for details.
    202. 

  202. source_ckpt_folder_name:    folder name to load the checkpoint from (self.experiment_name if none is given)
    203. 

  203. load_weights_only:          loads only the weight from the checkpoint and zeroize the training params
    204. 

  204. load_backbone:              loads the provided checkpoint to self.net.backbone instead of self.net
    205. 

  205. external_checkpoint_path:   The path to the external checkpoint to be loaded. Can be absolute or relative</p>
    206. 

  206. <blockquote>
    207. 

  207. <div><p>(ie: path/to/checkpoint.pth). If provided, will automatically attempt to
    208. 

  208. load the checkpoint even if the load_checkpoint flag is not provided.</p>
    209. 

  209. </div></blockquote>
    210. 

  210. </p></li>
    211. 

  211. </ul>
    212. 

  212. </dd>
    213. 

  213. </dl>
    214. 

  214. </dd></dl>
    215. 

  215. 

  216. <dl class="py method">
    217. 

  217. <dt class="sig sig-object py" id="id0">
    218. 

  218. <span class="sig-name descname"><span class="pre">train</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">training_params</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</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/sg_model/sg_model.html#SgModel.train"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#id0" title="Permalink to this definition"></a></dt>
    219. 

  219. <dd><p>train - Trains the Model</p>
    220. 

  220. <dl>
    221. 

  221. <dt>IMPORTANT NOTE: Additional batch parameters can be added as a third item (optional) if a tuple is returned by</dt><dd><p>the data loaders, as dictionary. The phase context will hold the additional items, under an attribute with
    222. 

  222. the same name as the key in this dictionary. Then such items can be accessed through phase callbacks.</p>
    223. 

  223. <blockquote>
    224. 

  224. <div><dl class="field-list">
    225. 

  225. <dt class="field-odd">param training_params</dt>
    226. 

  226. <dd class="field-odd"><ul>
    227. 

  227. <li><p><cite>max_epochs</cite> : int</p>
    228. 

  228. <blockquote>
    229. 

  229. <div><p>Number of epochs to run training.</p>
    230. 

  230. </div></blockquote>
    231. 

  231. </li>
    232. 

  232. <li><p><cite>lr_updates</cite> : list(int)</p>
    233. 

  233. <blockquote>
    234. 

  234. <div><p>List of fixed epoch numbers to perform learning rate updates when <cite>lr_mode=’step’</cite>.</p>
    235. 

  235. </div></blockquote>
    236. 

  236. </li>
    237. 

  237. <li><p><cite>lr_decay_factor</cite> : float</p>
    238. 

  238. <blockquote>
    239. 

  239. <div><p>Decay factor to apply to the learning rate at each update when <cite>lr_mode=’step’</cite>.</p>
    240. 

  240. </div></blockquote>
    241. 

  241. </li>
    242. 

  242. <li><p><cite>lr_mode</cite> : str</p>
    243. 

  243. <blockquote>
    244. 

  244. <div><p>Learning rate scheduling policy, one of [‘step’,’poly’,’cosine’,’function’]. ‘step’ refers to
    245. 

  245. constant updates at epoch numbers passed through <cite>lr_updates</cite>. ‘cosine’ refers to Cosine Anealing
    246. 

  246. policy as mentioned in <a class="reference external" href="https://arxiv.org/abs/1608.03983">https://arxiv.org/abs/1608.03983</a>. ‘poly’ refers to polynomial decrease i.e
    247. 

  247. in each epoch iteration <cite>self.lr = self.initial_lr * pow((1.0 - (current_iter / max_iter)),
    248. 

  248. 0.9)</cite> ‘function’ refers to user defined learning rate scheduling function, that is passed through
    249. 

  249. <cite>lr_schedule_function</cite>.</p>
    250. 

  250. </div></blockquote>
    251. 

  251. </li>
    252. 

  252. <li><p><cite>lr_schedule_function</cite> : Union[callable,None]</p>
    253. 

  253. <blockquote>
    254. 

  254. <div><p>Learning rate scheduling function to be used when <cite>lr_mode</cite> is ‘function’.</p>
    255. 

  255. </div></blockquote>
    256. 

  256. </li>
    257. 

  257. <li><p><cite>lr_warmup_epochs</cite> : int (default=0)</p>
    258. 

  258. <blockquote>
    259. 

  259. <div><p>Number of epochs for learning rate warm up - see <a class="reference external" href="https://arxiv.org/pdf/1706.02677.pdf">https://arxiv.org/pdf/1706.02677.pdf</a> (Section 2.2).</p>
    260. 

  260. </div></blockquote>
    261. 

  261. </li>
    262. 

  262. <li><dl class="simple">
    263. 

  263. <dt><cite>cosine_final_lr_ratio</cite><span class="classifier">float (default=0.01)</span></dt><dd><dl class="simple">
    264. 

  264. <dt>Final learning rate ratio (only relevant when <a href="#id1"><span class="problematic" id="id2">`</span></a>lr_mode`=’cosine’). The cosine starts from initial_lr and reaches</dt><dd><p>initial_lr * cosine_final_lr_ratio in last epoch</p>
    265. 

  265. </dd>
    266. 

  266. </dl>
    267. 

  267. </dd>
    268. 

  268. </dl>
    269. 

  269. </li>
    270. 

  270. <li><p><cite>inital_lr</cite> : float</p>
    271. 

  271. <blockquote>
    272. 

  272. <div><p>Initial learning rate.</p>
    273. 

  273. </div></blockquote>
    274. 

  274. </li>
    275. 

  275. <li><p><cite>loss</cite> : Union[nn.module, str]</p>
    276. 

  276. <blockquote>
    277. 

  277. <div><p>Loss function for training.
    278. 

  278. One of SuperGradient’s built in options:</p>
    279. 

  279. <blockquote>
    280. 

  280. <div><p>“cross_entropy”: LabelSmoothingCrossEntropyLoss,
    281. 

  281. “mse”: MSELoss,
    282. 

  282. “r_squared_loss”: RSquaredLoss,
    283. 

  283. “detection_loss”: YoLoV3DetectionLoss,
    284. 

  284. “shelfnet_ohem_loss”: ShelfNetOHEMLoss,
    285. 

  285. “shelfnet_se_loss”: ShelfNetSemanticEncodingLoss,
    286. 

  286. “ssd_loss”: SSDLoss,</p>
    287. 

  287. </div></blockquote>
    288. 

  288. <p>or user defined nn.module loss function.</p>
    289. 

  289. <p>IMPORTANT: forward(…) should return a (loss, loss_items) tuple where loss is the tensor used
    290. 

  290. for backprop (i.e what your original loss function returns), and loss_items should be a tensor of
    291. 

  291. shape (n_items), of values computed during the forward pass which we desire to log over the
    292. 

  292. entire epoch. For example- the loss itself should always be logged. Another example is a scenario
    293. 

  293. where the computed loss is the sum of a few components we would like to log- these entries in
    294. 

  294. loss_items).</p>
    295. 

  295. <p>When training, set the loss_logging_items_names parameter in train_params to be a list of
    296. 

  296. strings, of length n_items who’s ith element is the name of the ith entry in loss_items. Then
    297. 

  297. each item will be logged, rendered on tensorboard and “watched” (i.e saving model checkpoints
    298. 

  298. according to it).</p>
    299. 

  299. <p>Since running logs will save the loss_items in some internal state, it is recommended that
    300. 

  300. loss_items are detached from their computational graph for memory efficiency.</p>
    301. 

  301. </div></blockquote>
    302. 

  302. </li>
    303. 

  303. <li><p><cite>optimizer</cite> : Union[str, torch.optim.Optimizer]</p>
    304. 

  304. <blockquote>
    305. 

  305. <div><p>Optimization algorithm. One of [‘Adam’,’SGD’,’RMSProp’] corresponding to the torch.optim
    306. 

  306. optimzers implementations, or any object that implements torch.optim.Optimizer.</p>
    307. 

  307. </div></blockquote>
    308. 

  308. </li>
    309. 

  309. <li><p><cite>criterion_params</cite> : dict</p>
    310. 

  310. <blockquote>
    311. 

  311. <div><p>Loss function parameters.</p>
    312. 

  312. </div></blockquote>
    313. 

  313. </li>
    314. 

  314. <li><dl>
    315. 

  315. <dt><cite>optimizer_params</cite><span class="classifier">dict</span></dt><dd><p>When <cite>optimizer</cite> is one of [‘Adam’,’SGD’,’RMSProp’], it will be initialized with optimizer_params.</p>
    316. 

  316. <p>(see <a class="reference external" href="https://pytorch.org/docs/stable/optim.html">https://pytorch.org/docs/stable/optim.html</a> for the full list of
    317. 

  317. parameters for each optimizer).</p>
    318. 

  318. </dd>
    319. 

  319. </dl>
    320. 

  320. </li>
    321. 

  321. <li><p><cite>train_metrics_list</cite> : list(torchmetrics.Metric)</p>
    322. 

  322. <blockquote>
    323. 

  323. <div><p>Metrics to log during training. For more information on torchmetrics see
    324. 

  324. <a class="reference external" href="https://torchmetrics.rtfd.io/en/latest/">https://torchmetrics.rtfd.io/en/latest/</a>.</p>
    325. 

  325. </div></blockquote>
    326. 

  326. </li>
    327. 

  327. <li><p><cite>valid_metrics_list</cite> : list(torchmetrics.Metric)</p>
    328. 

  328. <blockquote>
    329. 

  329. <div><p>Metrics to log during validation/testing. For more information on torchmetrics see
    330. 

  330. <a class="reference external" href="https://torchmetrics.rtfd.io/en/latest/">https://torchmetrics.rtfd.io/en/latest/</a>.</p>
    331. 

  331. </div></blockquote>
    332. 

  332. </li>
    333. 

  333. <li><p><cite>loss_logging_items_names</cite> : list(str)</p>
    334. 

  334. <blockquote>
    335. 

  335. <div><p>The list of names/titles for the outputs returned from the loss functions forward pass (reminder-
    336. 

  336. the loss function should return the tuple (loss, loss_items)). These names will be used for
    337. 

  337. logging their values.</p>
    338. 

  338. </div></blockquote>
    339. 

  339. </li>
    340. 

  340. <li><p><cite>metric_to_watch</cite> : str (default=”Accuracy”)</p>
    341. 

  341. <blockquote>
    342. 

  342. <div><p>will be the metric which the model checkpoint will be saved according to, and can be set to any
    343. 

  343. of the following:</p>
    344. 

  344. <blockquote>
    345. 

  345. <div><p>a metric name (str) of one of the metric objects from the valid_metrics_list</p>
    346. 

  346. <p>a “metric_name” if some metric in valid_metrics_list has an attribute component_names which
    347. 

  347. is a list referring to the names of each entry in the output metric (torch tensor of size n)</p>
    348. 

  348. <p>one of “loss_logging_items_names” i.e which will correspond to an item returned during the
    349. 

  349. loss function’s forward pass.</p>
    350. 

  350. </div></blockquote>
    351. 

  351. <p>At the end of each epoch, if a new best metric_to_watch value is achieved, the models checkpoint
    352. 

  352. is saved in YOUR_PYTHON_PATH/checkpoints/ckpt_best.pth</p>
    353. 

  353. </div></blockquote>
    354. 

  354. </li>
    355. 

  355. <li><p><cite>greater_metric_to_watch_is_better</cite> : bool</p>
    356. 

  356. <blockquote>
    357. 

  357. <div><dl class="simple">
    358. 

  358. <dt>When choosing a model’s checkpoint to be saved, the best achieved model is the one that maximizes the</dt><dd><p>metric_to_watch when this parameter is set to True, and a one that minimizes it otherwise.</p>
    359. 

  359. </dd>
    360. 

  360. </dl>
    361. 

  361. </div></blockquote>
    362. 

  362. </li>
    363. 

  363. <li><p><cite>ema</cite> : bool (default=False)</p>
    364. 

  364. <blockquote>
    365. 

  365. <div><p>Whether to use Model Exponential Moving Average (see
    366. 

  366. <a class="reference external" href="https://github.com/rwightman/pytorch-image-models">https://github.com/rwightman/pytorch-image-models</a> ema implementation)</p>
    367. 

  367. </div></blockquote>
    368. 

  368. </li>
    369. 

  369. <li><p><cite>batch_accumulate</cite> : int (default=1)</p>
    370. 

  370. <blockquote>
    371. 

  371. <div><p>Number of batches to accumulate before every backward pass.</p>
    372. 

  372. </div></blockquote>
    373. 

  373. </li>
    374. 

  374. <li><p><cite>ema_params</cite> : dict</p>
    375. 

  375. <blockquote>
    376. 

  376. <div><p>Parameters for the ema model.</p>
    377. 

  377. </div></blockquote>
    378. 

  378. </li>
    379. 

  379. <li><p><cite>zero_weight_decay_on_bias_and_bn</cite> : bool (default=False)</p>
    380. 

  380. <blockquote>
    381. 

  381. <div><p>Whether to apply weight decay on batch normalization parameters or not (ignored when the passed
    382. 

  382. optimizer has already been initialized).</p>
    383. 

  383. </div></blockquote>
    384. 

  384. </li>
    385. 

  385. <li><p><cite>load_opt_params</cite> : bool (default=True)</p>
    386. 

  386. <blockquote>
    387. 

  387. <div><p>Whether to load the optimizers parameters as well when loading a model’s checkpoint.</p>
    388. 

  388. </div></blockquote>
    389. 

  389. </li>
    390. 

  390. <li><p><cite>run_validation_freq</cite> : int (default=1)</p>
    391. 

  391. <blockquote>
    392. 

  392. <div><dl class="simple">
    393. 

  393. <dt>The frequency in which validation is performed during training (i.e the validation is ran every</dt><dd><p><cite>run_validation_freq</cite> epochs.</p>
    394. 

  394. </dd>
    395. 

  395. </dl>
    396. 

  396. </div></blockquote>
    397. 

  397. </li>
    398. 

  398. <li><p><cite>save_model</cite> : bool (default=True)</p>
    399. 

  399. <blockquote>
    400. 

  400. <div><p>Whether to save the model checkpoints.</p>
    401. 

  401. </div></blockquote>
    402. 

  402. </li>
    403. 

  403. <li><p><cite>silent_mode</cite> : bool</p>
    404. 

  404. <blockquote>
    405. 

  405. <div><p>Silents the print outs.</p>
    406. 

  406. </div></blockquote>
    407. 

  407. </li>
    408. 

  408. <li><p><cite>mixed_precision</cite> : bool</p>
    409. 

  409. <blockquote>
    410. 

  410. <div><p>Whether to use mixed precision or not.</p>
    411. 

  411. </div></blockquote>
    412. 

  412. </li>
    413. 

  413. <li><p><cite>save_ckpt_epoch_list</cite> : list(int) (default=[])</p>
    414. 

  414. <blockquote>
    415. 

  415. <div><p>List of fixed epoch indices the user wishes to save checkpoints in.</p>
    416. 

  416. </div></blockquote>
    417. 

  417. </li>
    418. 

  418. <li><p><cite>average_best_models</cite> : bool (default=False)</p>
    419. 

  419. <blockquote>
    420. 

  420. <div><p>If set, a snapshot dictionary file and the average model will be saved / updated at every epoch
    421. 

  421. and evaluated only when training is completed. The snapshot file will only be deleted upon
    422. 

  422. completing the training. The snapshot dict will be managed on cpu.</p>
    423. 

  423. </div></blockquote>
    424. 

  424. </li>
    425. 

  425. <li><p><cite>precise_bn</cite> : bool (default=False)</p>
    426. 

  426. <blockquote>
    427. 

  427. <div><p>Whether to use precise_bn calculation during the training.</p>
    428. 

  428. </div></blockquote>
    429. 

  429. </li>
    430. 

  430. <li><p><cite>precise_bn_batch_size</cite> : int (default=None)</p>
    431. 

  431. <blockquote>
    432. 

  432. <div><p>The effective batch size we want to calculate the batchnorm on. For example, if we are training a model
    433. 

  433. on 8 gpus, with a batch of 128 on each gpu, a good rule of thumb would be to give it 8192
    434. 

  434. (ie: effective_batch_size * num_gpus = batch_per_gpu * num_gpus * num_gpus).
    435. 

  435. If precise_bn_batch_size is not provided in the training_params, the latter heuristic will be taken.</p>
    436. 

  436. </div></blockquote>
    437. 

  437. </li>
    438. 

  438. <li><p><cite>seed</cite> : int (default=42)</p>
    439. 

  439. <blockquote>
    440. 

  440. <div><p>Random seed to be set for torch, numpy, and random. When using DDP each process will have it’s seed
    441. 

  441. set to seed + rank.</p>
    442. 

  442. </div></blockquote>
    443. 

  443. </li>
    444. 

  444. <li><p><cite>log_installed_packages</cite> : bool (default=False)</p>
    445. 

  445. <blockquote>
    446. 

  446. <div><dl class="simple">
    447. 

  447. <dt>When set, the list of all installed packages (and their versions) will be written to the tensorboard</dt><dd><p>and logfile (useful when trying to reproduce results).</p>
    448. 

  448. </dd>
    449. 

  449. </dl>
    450. 

  450. </div></blockquote>
    451. 

  451. </li>
    452. 

  452. <li><p><cite>dataset_statistics</cite> : bool (default=False)</p>
    453. 

  453. <blockquote>
    454. 

  454. <div><p>Enable a statistic analysis of the dataset. If set to True the dataset will be analyzed and a report
    455. 

  455. will be added to the tensorboard along with some sample images from the dataset. Currently only
    456. 

  456. detection datasets are supported for analysis.</p>
    457. 

  457. </div></blockquote>
    458. 

  458. </li>
    459. 

  459. <li><p><cite>save_full_train_log</cite> : bool (default=False)</p>
    460. 

  460. <blockquote>
    461. 

  461. <div><dl class="simple">
    462. 

  462. <dt>When set, a full log (of all super_gradients modules, including uncaught exceptions from any other</dt><dd><p>module) of the training will be saved in the checkpoint directory under full_train_log.log</p>
    463. 

  463. </dd>
    464. 

  464. </dl>
    465. 

  465. </div></blockquote>
    466. 

  466. </li>
    467. 

  467. <li><p><cite>sg_logger</cite> : Union[AbstractSGLogger, str] (defauls=base_sg_logger)</p>
    468. 

  468. <blockquote>
    469. 

  469. <div><p>Define the SGLogger object for this training process. The SGLogger handles all disk writes, logs, TensorBoard, remote logging
    470. 

  470. and remote storage. By overriding the default base_sg_logger, you can change the storage location, support external monitoring and logging
    471. 

  471. or support remote storage.</p>
    472. 

  472. </div></blockquote>
    473. 

  473. </li>
    474. 

  474. <li><p><cite>sg_logger_params</cite> : dict</p>
    475. 

  475. <p>SGLogger parameters</p>
    476. 

  476. </li>
    477. 

  477. <li><p><cite>clip_grad_norm</cite> : float</p>
    478. 

  478. <p>Defines a maximal L2 norm of the gradients. Values which exceed the given value will be clipped</p>
    479. 

  479. </li>
    480. 

  480. <li><p><cite>lr_cooldown_epochs</cite> : int (default=0)</p>
    481. 

  481. <p>Number of epochs to cooldown LR (i.e the last epoch from scheduling view point=max_epochs-cooldown).</p>
    482. 

  482. </li>
    483. 

  483. <li><p><cite>pre_prediction_callback</cite> : Callable (default=None)</p>
    484. 

  484. <blockquote>
    485. 

  485. <div><dl class="simple">
    486. 

  486. <dt>When not None, this callback will be applied to images and targets, and returning them to be used</dt><dd><p>for the forward pass, and further computations. Args for this callable should be in the order
    487. 

  487. (inputs, targets, batch_idx) returning modified_inputs, modified_targets</p>
    488. 

  488. </dd>
    489. 

  489. </dl>
    490. 

  490. </div></blockquote>
    491. 

  491. </li>
    492. 

  492. <li><p><cite>ckpt_best_name</cite> : str (default=’ckpt_best.pth’)</p>
    493. 

  493. <p>The best checkpoint (according to metric_to_watch) will be saved under this filename in the checkpoints directory.</p>
    494. 

  494. </li>
    495. 

  495. <li><p><cite>enable_qat</cite>: bool (default=False)</p>
    496. 

  496. <dl class="simple">
    497. 

  497. <dt>Adds a QATCallback to the phase callbacks, that triggers quantization aware training starting from</dt><dd><p>qat_params[“start_epoch”]</p>
    498. 

  498. </dd>
    499. 

  499. </dl>
    500. 

  500. </li>
    501. 

  501. <li><p><cite>qat_params</cite>: dict-like object with the following key/values:</p>
    502. 

  502. <blockquote>
    503. 

  503. <div><p>start_epoch: int, first epoch to start QAT.</p>
    504. 

  504. <dl class="simple">
    505. 

  505. <dt>quant_modules_calib_method: str, One of [percentile, mse, entropy, max]. Statistics method for amax</dt><dd><p>computation of the quantized modules (default=percentile).</p>
    506. 

  506. </dd>
    507. 

  507. </dl>
    508. 

  508. <p>per_channel_quant_modules: bool, whether quant modules should be per channel (default=False).</p>
    509. 

  509. <p>calibrate: bool, whether to perfrom calibration (default=False).</p>
    510. 

  510. <p>calibrated_model_path: str, path to a calibrated checkpoint (default=None).</p>
    511. 

  511. <dl class="simple">
    512. 

  512. <dt>calib_data_loader: torch.utils.data.DataLoader, data loader of the calibration dataset. When None,</dt><dd><p>context.train_loader will be used (default=None).</p>
    513. 

  513. </dd>
    514. 

  514. </dl>
    515. 

  515. <p>num_calib_batches: int, number of batches to collect the statistics from.</p>
    516. 

  516. <dl class="simple">
    517. 

  517. <dt>percentile: float, percentile value to use when SgModel,quant_modules_calib_method=’percentile’.</dt><dd><p>Discarded when other methods are used (Default=99.99).</p>
    518. 

  518. </dd>
    519. 

  519. </dl>
    520. 

  520. </div></blockquote>
    521. 

  521. </li>
    522. 

  522. </ul>
    523. 

  523. </dd>
    524. 

  524. </dl>
    525. 

  525. </div></blockquote>
    526. 

  526. </dd>
    527. 

  527. </dl>
    528. 

  528. <dl class="field-list simple">
    529. 

  529. <dt class="field-odd">Returns</dt>
    530. 

  530. <dd class="field-odd"><p></p>
    531. 

  531. </dd>
    532. 

  532. </dl>
    533. 

  533. </dd></dl>
    534. 

  534. 

  535. <dl class="py method">
    536. 

  536. <dt class="sig sig-object py" id="id3">
    537. 

  537. <span class="sig-name descname"><span class="pre">predict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">inputs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</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">half</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">normalize</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">verbose</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">move_outputs_to_cpu</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/sg_model/sg_model.html#SgModel.predict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#id3" title="Permalink to this definition"></a></dt>
    538. 

  538. <dd><p>A fast predictor for a batch of inputs
    539. 

  539. :param inputs: torch.tensor or numpy.array</p>
    540. 

  540. <blockquote>
    541. 

  541. <div><p>a batch of inputs</p>
    542. 

  542. </div></blockquote>
    543. 

  543. <dl class="field-list simple">
    544. 

  544. <dt class="field-odd">Parameters</dt>
    545. 

  545. <dd class="field-odd"><ul class="simple">
    546. 

  546. <li><p><strong>targets</strong> – torch.tensor()
    547. 

  547. corresponding labels - if non are given - accuracy will not be computed</p></li>
    548. 

  548. <li><p><strong>verbose</strong> – bool
    549. 

  549. print the results to screen</p></li>
    550. 

  550. <li><p><strong>normalize</strong> – bool
    551. 

  551. If true, normalizes the tensor according to the dataloader’s normalization values</p></li>
    552. 

  552. <li><p><strong>half</strong> – Performs half precision evaluation</p></li>
    553. 

  553. <li><p><strong>move_outputs_to_cpu</strong> – Moves the results from the GPU to the CPU</p></li>
    554. 

  554. </ul>
    555. 

  555. </dd>
    556. 

  556. <dt class="field-even">Returns</dt>
    557. 

  557. <dd class="field-even"><p>outputs, acc, net_time, gross_time
    558. 

  558. networks predictions, accuracy calculation, forward pass net time, function gross time</p>
    559. 

  559. </dd>
    560. 

  560. </dl>
    561. 

  561. </dd></dl>
    562. 

  562. 

  563. <dl class="py property">
    564. 

  564. <dt class="sig sig-object py" id="super_gradients.training.SgModel.get_arch_params">
    565. 

  565. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_arch_params</span></span><a class="headerlink" href="#super_gradients.training.SgModel.get_arch_params" title="Permalink to this definition"></a></dt>
    566. 

  566. <dd></dd></dl>
    567. 

  567. 

  568. <dl class="py property">
    569. 

  569. <dt class="sig sig-object py" id="super_gradients.training.SgModel.get_structure">
    570. 

  570. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_structure</span></span><a class="headerlink" href="#super_gradients.training.SgModel.get_structure" title="Permalink to this definition"></a></dt>
    571. 

  571. <dd></dd></dl>
    572. 

  572. 

  573. <dl class="py property">
    574. 

  574. <dt class="sig sig-object py" id="super_gradients.training.SgModel.get_architecture">
    575. 

  575. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_architecture</span></span><a class="headerlink" href="#super_gradients.training.SgModel.get_architecture" title="Permalink to this definition"></a></dt>
    576. 

  576. <dd></dd></dl>
    577. 

  577. 

  578. <dl class="py method">
    579. 

  579. <dt class="sig sig-object py" id="super_gradients.training.SgModel.set_experiment_name">
    580. 

  580. <span class="sig-name descname"><span class="pre">set_experiment_name</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">experiment_name</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_experiment_name"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.set_experiment_name" title="Permalink to this definition"></a></dt>
    581. 

  581. <dd></dd></dl>
    582. 

  582. 

  583. <dl class="py property">
    584. 

  584. <dt class="sig sig-object py" id="super_gradients.training.SgModel.get_module">
    585. 

  585. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_module</span></span><a class="headerlink" href="#super_gradients.training.SgModel.get_module" title="Permalink to this definition"></a></dt>
    586. 

  586. <dd></dd></dl>
    587. 

  587. 

  588. <dl class="py method">
    589. 

  589. <dt class="sig sig-object py" id="super_gradients.training.SgModel.set_module">
    590. 

  590. <span class="sig-name descname"><span class="pre">set_module</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/sg_model/sg_model.html#SgModel.set_module"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.set_module" title="Permalink to this definition"></a></dt>
    591. 

  591. <dd></dd></dl>
    592. 

  592. 

  593. <dl class="py method">
    594. 

  594. <dt class="sig sig-object py" id="super_gradients.training.SgModel.test">
    595. 

  595. <span class="sig-name descname"><span class="pre">test</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">test_loader</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">torch.utils.data.dataloader.DataLoader</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">loss</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">torch.nn.modules.loss._Loss</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">silent_mode</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">test_metrics_list</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_logging_items_names</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_progress_verbose</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">test_phase_callbacks</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_ema_net</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> &#x2192; <span class="pre">tuple</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.test"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.test" title="Permalink to this definition"></a></dt>
    596. 

  596. <dd><p>Evaluates the model on given dataloader and metrics.</p>
    597. 

  597. <dl class="field-list simple">
    598. 

  598. <dt class="field-odd">Parameters</dt>
    599. 

  599. <dd class="field-odd"><ul class="simple">
    600. 

  600. <li><p><strong>test_loader</strong> – dataloader to perform test on.</p></li>
    601. 

  601. <li><p><strong>test_metrics_list</strong> – (list(torchmetrics.Metric)) metrics list for evaluation.</p></li>
    602. 

  602. <li><p><strong>silent_mode</strong> – (bool) controls verbosity</p></li>
    603. 

  603. <li><p><strong>metrics_progress_verbose</strong> – (bool) controls the verbosity of metrics progress (default=False). Slows down the program.</p></li>
    604. 

  604. </ul>
    605. 

  605. </dd>
    606. 

  606. </dl>
    607. 

  607. <dl class="simple">
    608. 

  608. <dt>:param use_ema_net (bool) whether to perform test on self.ema_model.ema (when self.ema_model.ema exists,</dt><dd><p>otherwise self.net will be tested) (default=True)</p>
    609. 

  609. </dd>
    610. 

  610. </dl>
    611. 

  611. <dl class="field-list simple">
    612. 

  612. <dt class="field-odd">Returns</dt>
    613. 

  613. <dd class="field-odd"><p>results tuple (tuple) containing the loss items and metric values.</p>
    614. 

  614. </dd>
    615. 

  615. </dl>
    616. 

  616. <dl class="simple">
    617. 

  617. <dt>All of the above args will override SgModel’s corresponding attribute when not equal to None. Then evaluation</dt><dd><p>is ran on self.test_loader with self.test_metrics.</p>
    618. 

  618. </dd>
    619. 

  619. </dl>
    620. 

  620. </dd></dl>
    621. 

  621. 

  622. <dl class="py method">
    623. 

  623. <dt class="sig sig-object py" id="super_gradients.training.SgModel.evaluate">
    624. 

  624. <span class="sig-name descname"><span class="pre">evaluate</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data_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">metrics</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torchmetrics.collections.MetricCollection</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">evaluation_type</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.common.data_types.enum.html#super_gradients.common.data_types.enum.EvaluationType" title="super_gradients.common.data_types.enum.evaluation_type.EvaluationType"><span class="pre">super_gradients.common.data_types.enum.evaluation_type.EvaluationType</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">epoch</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">silent_mode</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">metrics_progress_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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.evaluate"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.evaluate" title="Permalink to this definition"></a></dt>
    625. 

  625. <dd><p>Evaluates the model on given dataloader and metrics.</p>
    626. 

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

  627. <dt class="field-odd">Parameters</dt>
    628. 

  628. <dd class="field-odd"><ul class="simple">
    629. 

  629. <li><p><strong>data_loader</strong> – dataloader to perform evaluataion on</p></li>
    630. 

  630. <li><p><strong>metrics</strong> – (MetricCollection) metrics for evaluation</p></li>
    631. 

  631. <li><p><strong>evaluation_type</strong> – (EvaluationType) controls which phase callbacks will be used (for example, on batch end,
    632. 

  632. when evaluation_type=EvaluationType.VALIDATION the Phase.VALIDATION_BATCH_END callbacks will be triggered)</p></li>
    633. 

  633. <li><p><strong>epoch</strong> – (int) epoch idx</p></li>
    634. 

  634. <li><p><strong>silent_mode</strong> – (bool) controls verbosity</p></li>
    635. 

  635. <li><p><strong>metrics_progress_verbose</strong> – (bool) controls the verbosity of metrics progress (default=False).
    636. 

  636. Slows down the program significantly.</p></li>
    637. 

  637. </ul>
    638. 

  638. </dd>
    639. 

  639. <dt class="field-even">Returns</dt>
    640. 

  640. <dd class="field-even"><p>results tuple (tuple) containing the loss items and metric values.</p>
    641. 

  641. </dd>
    642. 

  642. </dl>
    643. 

  643. </dd></dl>
    644. 

  644. 

  645. <dl class="py property">
    646. 

  646. <dt class="sig sig-object py" id="super_gradients.training.SgModel.get_net">
    647. 

  647. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_net</span></span><a class="headerlink" href="#super_gradients.training.SgModel.get_net" title="Permalink to this definition"></a></dt>
    648. 

  648. <dd><p>Getter for network.
    649. 

  649. :return: torch.nn.Module, self.net</p>
    650. 

  650. </dd></dl>
    651. 

  651. 

  652. <dl class="py method">
    653. 

  653. <dt class="sig sig-object py" id="super_gradients.training.SgModel.set_net">
    654. 

  654. <span class="sig-name descname"><span class="pre">set_net</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_net"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.set_net" title="Permalink to this definition"></a></dt>
    655. 

  655. <dd><p>Setter for network.</p>
    656. 

  656. <dl class="field-list simple">
    657. 

  657. <dt class="field-odd">Parameters</dt>
    658. 

  658. <dd class="field-odd"><p><strong>net</strong> – torch.nn.Module, value to set net</p>
    659. 

  659. </dd>
    660. 

  660. <dt class="field-even">Returns</dt>
    661. 

  661. <dd class="field-even"><p></p>
    662. 

  662. </dd>
    663. 

  663. </dl>
    664. 

  664. </dd></dl>
    665. 

  665. 

  666. <dl class="py method">
    667. 

  667. <dt class="sig sig-object py" id="super_gradients.training.SgModel.set_ckpt_best_name">
    668. 

  668. <span class="sig-name descname"><span class="pre">set_ckpt_best_name</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ckpt_best_name</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_ckpt_best_name"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.set_ckpt_best_name" title="Permalink to this definition"></a></dt>
    669. 

  669. <dd><p>Setter for best checkpoint filename.</p>
    670. 

  670. <dl class="field-list simple">
    671. 

  671. <dt class="field-odd">Parameters</dt>
    672. 

  672. <dd class="field-odd"><p><strong>ckpt_best_name</strong> – str, value to set ckpt_best_name</p>
    673. 

  673. </dd>
    674. 

  674. </dl>
    675. 

  675. </dd></dl>
    676. 

  676. 

  677. <dl class="py method">
    678. 

  678. <dt class="sig sig-object py" id="super_gradients.training.SgModel.set_ema">
    679. 

  679. <span class="sig-name descname"><span class="pre">set_ema</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">val</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/sg_model/sg_model.html#SgModel.set_ema"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SgModel.set_ema" title="Permalink to this definition"></a></dt>
    680. 

  680. <dd><p>Setter for self.ema</p>
    681. 

  681. <dl class="field-list simple">
    682. 

  682. <dt class="field-odd">Parameters</dt>
    683. 

  683. <dd class="field-odd"><p><strong>val</strong> – bool, value to set ema</p>
    684. 

  684. </dd>
    685. 

  685. </dl>
    686. 

  686. </dd></dl>
    687. 

  687. 

  688. </dd></dl>
    689. 

  689. 

  690. <dl class="py class">
    691. 

  691. <dt class="sig sig-object py" id="super_gradients.training.KDModel">
    692. 

  692. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">KDModel</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">args</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/kd_model/kd_model.html#KDModel"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.KDModel" title="Permalink to this definition"></a></dt>
    693. 

  693. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.sg_model.html#super_gradients.training.sg_model.sg_model.SgModel" title="super_gradients.training.sg_model.sg_model.SgModel"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.sg_model.sg_model.SgModel</span></code></a></p>
    694. 

  694. <dl class="py method">
    695. 

  695. <dt class="sig sig-object py" id="super_gradients.training.KDModel.build_model">
    696. 

  696. <span class="sig-name descname"><span class="pre">build_model</span></span><span class="sig-paren">(</span><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">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">super_gradients.training.models.kd_modules.kd_module.KDModule</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">'kd_module'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">arch_params</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">checkpoint_params</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="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</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/kd_model/kd_model.html#KDModel.build_model"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.KDModel.build_model" title="Permalink to this definition"></a></dt>
    697. 

  697. <dd><dl class="field-list simple">
    698. 

  698. <dt class="field-odd">Parameters</dt>
    699. 

  699. <dd class="field-odd"><ul class="simple">
    700. 

  700. <li><p><strong>architecture</strong> – (Union[str, KDModule]) Defines the network’s architecture from models/KD_ARCHITECTURES
    701. 

  701. (default=’kd_module’)</p></li>
    702. 

  702. <li><p><strong>arch_params</strong> – (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc to be passed to kd
    703. 

  703. architecture class (discarded when architecture is KDModule instance)</p></li>
    704. 

  704. <li><p><strong>checkpoint_params</strong> – <p>(dict) A dictionary like object with the following keys/values:</p>
    705. 

  705. <p>student_pretrained_weights:   String describing the dataset of the pretrained weights (for example
    706. 

  706. “imagenent”) for the student network.</p>
    707. 

  707. <p>teacher_pretrained_weights:   String describing the dataset of the pretrained weights (for example
    708. 

  708. “imagenent”) for the teacher network.</p>
    709. 

  709. <dl class="simple">
    710. 

  710. <dt>teacher_checkpoint_path:    Local path to the teacher’s checkpoint. Note that when passing pretrained_weights</dt><dd><p>through teacher_arch_params these weights will be overridden by the
    711. 

  711. pretrained checkpoint. (default=None)</p>
    712. 

  712. </dd>
    713. 

  713. <dt>load_kd_model_checkpoint:   Whether to load an entire KDModule checkpoint (used to continue KD training)</dt><dd><p>(default=False)</p>
    714. 

  714. </dd>
    715. 

  715. <dt>kd_model_source_ckpt_folder_name: Folder name to load an entire KDModule checkpoint from</dt><dd><p>(self.experiment_name if none is given) to resume KD training (default=None)</p>
    716. 

  716. </dd>
    717. 

  717. <dt>kd_model_external_checkpoint_path: The path to the external checkpoint to be loaded. Can be absolute or relative</dt><dd><p>(ie: path/to/checkpoint.pth). If provided, will automatically attempt to
    718. 

  718. load the checkpoint even if the load_checkpoint flag is not provided.
    719. 

  719. (deafult=None)</p>
    720. 

  720. </dd>
    721. 

  721. </dl>
    722. 

  722. </p></li>
    723. 

  723. </ul>
    724. 

  724. </dd>
    725. 

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

  726. <dd class="field-even"><ul class="simple">
    727. 

  727. <li><p><strong>student_architecture</strong> – (Union[str, SgModule]) Defines the student’s architecture from
    728. 

  728. models/ALL_ARCHITECTURES (when str), or directly defined the student network (when SgModule).</p></li>
    729. 

  729. <li><p><strong>teacher_architecture</strong> – (Union[str, SgModule]) Defines the teacher’s architecture from
    730. 

  730. models/ALL_ARCHITECTURES (when str), or directly defined the teacher network (when SgModule).</p></li>
    731. 

  731. <li><p><strong>student_arch_params</strong> – (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for student
    732. 

  732. net. (deafult={})</p></li>
    733. 

  733. <li><p><strong>teacher_arch_params</strong> – (dict) Architecture H.P. e.g.: block, num_blocks, num_classes, etc for teacher
    734. 

  734. net. (deafult={})</p></li>
    735. 

  735. <li><p><strong>run_teacher_on_eval</strong> – (bool)- whether to run self.teacher at eval mode regardless of self.train(mode)</p></li>
    736. 

  736. </ul>
    737. 

  737. </dd>
    738. 

  738. </dl>
    739. 

  739. </dd></dl>
    740. 

  740. 

  741. </dd></dl>
    742. 

  742. 

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

  744. <dt class="sig sig-object py" id="super_gradients.training.MultiGPUMode">
    745. 

  745. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">MultiGPUMode</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/common/data_types/enum/multi_gpu_mode.html#MultiGPUMode"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.MultiGPUMode" title="Permalink to this definition"></a></dt>
    746. 

  746. <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>
    747. 

  747. <dl class="py attribute">
    748. 

  748. <dt class="sig sig-object py">
    749. 

  749. <span class="sig-name descname"><span class="pre">OFF</span>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; <span class="pre">-</span> <span class="pre">Single</span> <span class="pre">GPU</span> <span class="pre">Mode</span> <span class="pre">/</span> <span class="pre">CPU</span> <span class="pre">Mode</span></span></dt>
    750. 

  750. <dd></dd></dl>
    751. 

  751. 

  752. <dl class="py attribute">
    753. 

  753. <dt class="sig sig-object py">
    754. 

  754. <span class="sig-name descname"><span class="pre">DATA_PARALLEL</span>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; <span class="pre">-</span> <span class="pre">Multiple</span> <span class="pre">GPUs,</span> <span class="pre">Synchronous</span></span></dt>
    755. 

  755. <dd></dd></dl>
    756. 

  756. 

  757. <dl class="py attribute">
    758. 

  758. <dt class="sig sig-object py">
    759. 

  759. <span class="sig-name descname"><span class="pre">DISTRIBUTED_DATA_PARALLEL</span> <span class="pre">-</span> <span class="pre">Multiple</span> <span class="pre">GPUs,</span> <span class="pre">Asynchronous</span></span></dt>
    760. 

  760. <dd></dd></dl>
    761. 

  761. 

  762. <dl class="py attribute">
    763. 

  763. <dt class="sig sig-object py" id="super_gradients.training.MultiGPUMode.OFF">
    764. 

  764. <span class="sig-name descname"><span class="pre">OFF</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'Off'</span></em><a class="headerlink" href="#super_gradients.training.MultiGPUMode.OFF" title="Permalink to this definition"></a></dt>
    765. 

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

  766. 

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

  768. <dt class="sig sig-object py" id="super_gradients.training.MultiGPUMode.DATA_PARALLEL">
    769. 

  769. <span class="sig-name descname"><span class="pre">DATA_PARALLEL</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'DP'</span></em><a class="headerlink" href="#super_gradients.training.MultiGPUMode.DATA_PARALLEL" title="Permalink to this definition"></a></dt>
    770. 

  770. <dd></dd></dl>
    771. 

  771. 

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

  773. <dt class="sig sig-object py" id="super_gradients.training.MultiGPUMode.DISTRIBUTED_DATA_PARALLEL">
    774. 

  774. <span class="sig-name descname"><span class="pre">DISTRIBUTED_DATA_PARALLEL</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'DDP'</span></em><a class="headerlink" href="#super_gradients.training.MultiGPUMode.DISTRIBUTED_DATA_PARALLEL" 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.MultiGPUMode.AUTO">
    779. 

  779. <span class="sig-name descname"><span class="pre">AUTO</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'AUTO'</span></em><a class="headerlink" href="#super_gradients.training.MultiGPUMode.AUTO" title="Permalink to this definition"></a></dt>
    780. 

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

  781. 

  782. </dd></dl>
    783. 

  783. 

  784. <dl class="py class">
    785. 

  785. <dt class="sig sig-object py" id="super_gradients.training.SegmentationTestDatasetInterface">
    786. 

  786. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">SegmentationTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">512</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">4</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#SegmentationTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.SegmentationTestDatasetInterface" title="Permalink to this definition"></a></dt>
    787. 

  787. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    788. 

  788. </dd></dl>
    789. 

  789. 

  790. <dl class="py class">
    791. 

  791. <dt class="sig sig-object py" id="super_gradients.training.DetectionTestDatasetInterface">
    792. 

  792. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">DetectionTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">320</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">4</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#DetectionTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.DetectionTestDatasetInterface" title="Permalink to this definition"></a></dt>
    793. 

  793. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    794. 

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

  795. 

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

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

  798. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">ClassificationTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">32</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">5</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#ClassificationTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.ClassificationTestDatasetInterface" title="Permalink to this definition"></a></dt>
    799. 

  799. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    800. 

  800. </dd></dl>
    801. 

  801. 

  802. <dl class="py class">
    803. 

  803. <dt class="sig sig-object py" id="super_gradients.training.StrictLoad">
    804. 

  804. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.</span></span><span class="sig-name descname"><span class="pre">StrictLoad</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/common/data_types/enum/strict_load.html#StrictLoad"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.StrictLoad" title="Permalink to this definition"></a></dt>
    805. 

  805. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">enum.Enum</span></code></p>
    806. 

  806. <dl>
    807. 

  807. <dt>Wrapper for adding more functionality to torch’s strict_load parameter in load_state_dict().</dt><dd><dl>
    808. 

  808. <dt>Attributes:</dt><dd><p>OFF              - Native torch “strict_load = off” behaviour. See nn.Module.load_state_dict() documentation for more details.
    809. 

  809. ON               - Native torch “strict_load = on” behaviour. See nn.Module.load_state_dict() documentation for more details.
    810. 

  810. NO_KEY_MATCHING  - Allows the usage of SuperGradient’s adapt_checkpoint function, which loads a checkpoint by matching each</p>
    811. 

  811. <blockquote>
    812. 

  812. <div><p>layer’s shapes (and bypasses the strict matching of the names of each layer (ie: disregards the state_dict key matching)).</p>
    813. 

  813. </div></blockquote>
    814. 

  814. </dd>
    815. 

  815. </dl>
    816. 

  816. </dd>
    817. 

  817. </dl>
    818. 

  818. <dl class="py attribute">
    819. 

  819. <dt class="sig sig-object py" id="super_gradients.training.StrictLoad.OFF">
    820. 

  820. <span class="sig-name descname"><span class="pre">OFF</span></span><em class="property"> <span class="pre">=</span> <span class="pre">False</span></em><a class="headerlink" href="#super_gradients.training.StrictLoad.OFF" title="Permalink to this definition"></a></dt>
    821. 

  821. <dd></dd></dl>
    822. 

  822. 

  823. <dl class="py attribute">
    824. 

  824. <dt class="sig sig-object py" id="super_gradients.training.StrictLoad.ON">
    825. 

  825. <span class="sig-name descname"><span class="pre">ON</span></span><em class="property"> <span class="pre">=</span> <span class="pre">True</span></em><a class="headerlink" href="#super_gradients.training.StrictLoad.ON" title="Permalink to this definition"></a></dt>
    826. 

  826. <dd></dd></dl>
    827. 

  827. 

  828. <dl class="py attribute">
    829. 

  829. <dt class="sig sig-object py" id="super_gradients.training.StrictLoad.NO_KEY_MATCHING">
    830. 

  830. <span class="sig-name descname"><span class="pre">NO_KEY_MATCHING</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'no_key_matching'</span></em><a class="headerlink" href="#super_gradients.training.StrictLoad.NO_KEY_MATCHING" title="Permalink to this definition"></a></dt>
    831. 

  831. <dd></dd></dl>
    832. 

  832. 

  833. </dd></dl>
    834. 

  834. 

  835. </section>
    836. 

  836. <section id="super-gradients-training-datasets-module">
    837. 

  837. <h2>super_gradients.training.datasets module<a class="headerlink" href="#super-gradients-training-datasets-module" title="Permalink to this headline"></a></h2>
    838. 

  838. <span class="target" id="module-super_gradients.training.datasets"></span><dl class="py class">
    839. 

  839. <dt class="sig sig-object py" id="super_gradients.training.datasets.DataAugmentation">
    840. 

  840. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">DataAugmentation</span></span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DataAugmentation" title="Permalink to this definition"></a></dt>
    841. 

  841. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    842. 

  842. <dl class="py method">
    843. 

  843. <dt class="sig sig-object py" id="super_gradients.training.datasets.DataAugmentation.to_tensor">
    844. 

  844. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">to_tensor</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.to_tensor"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DataAugmentation.to_tensor" title="Permalink to this definition"></a></dt>
    845. 

  845. <dd></dd></dl>
    846. 

  846. 

  847. <dl class="py method">
    848. 

  848. <dt class="sig sig-object py" id="super_gradients.training.datasets.DataAugmentation.normalize">
    849. 

  849. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">normalize</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">mean</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">std</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.normalize"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DataAugmentation.normalize" title="Permalink to this definition"></a></dt>
    850. 

  850. <dd></dd></dl>
    851. 

  851. 

  852. <dl class="py method">
    853. 

  853. <dt class="sig sig-object py" id="super_gradients.training.datasets.DataAugmentation.cutout">
    854. 

  854. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">cutout</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">mask_size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p</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">cutout_inside</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">mask_color</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">(0,</span> <span class="pre">0,</span> <span class="pre">0)</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/data_augmentation.html#DataAugmentation.cutout"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DataAugmentation.cutout" title="Permalink to this definition"></a></dt>
    855. 

  855. <dd></dd></dl>
    856. 

  856. 

  857. </dd></dl>
    858. 

  858. 

  859. <dl class="py class">
    860. 

  860. <dt class="sig sig-object py" id="super_gradients.training.datasets.ListDataset">
    861. 

  861. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">ListDataset</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">root</span></em>, <em class="sig-param"><span class="pre">file</span></em>, <em class="sig-param"><span class="pre">sample_loader:</span> <span class="pre">Callable</span> <span class="pre">=</span> <span class="pre">&lt;function</span> <span class="pre">default_loader&gt;</span></em>, <em class="sig-param"><span class="pre">target_loader:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">collate_fn:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">sample_extensions:</span> <span class="pre">tuple</span> <span class="pre">=</span> <span class="pre">('.jpg'</span></em>, <em class="sig-param"><span class="pre">'.jpeg'</span></em>, <em class="sig-param"><span class="pre">'.png'</span></em>, <em class="sig-param"><span class="pre">'.ppm'</span></em>, <em class="sig-param"><span class="pre">'.bmp'</span></em>, <em class="sig-param"><span class="pre">'.pgm'</span></em>, <em class="sig-param"><span class="pre">'.tif'</span></em>, <em class="sig-param"><span class="pre">'.tiff'</span></em>, <em class="sig-param"><span class="pre">'.webp')</span></em>, <em class="sig-param"><span class="pre">sample_transform:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">target_transform:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">target_extension='.npy'</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/sg_dataset.html#ListDataset"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.ListDataset" title="Permalink to this definition"></a></dt>
    862. 

  862. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    863. 

  863. <dl>
    864. 

  864. <dt>ListDataset - A PyTorch Vision Data Set extension that receives a file with FULL PATH to each of the samples.</dt><dd><p>Then, the assumption is that for every sample, there is a * matching target * in the same
    865. 

  865. path but with a different extension, i.e:</p>
    866. 

  866. <blockquote>
    867. 

  867. <div><dl class="simple">
    868. 

  868. <dt>for the samples paths:  (That appear in the list file)</dt><dd><p>/root/dataset/class_x/sample1.png
    869. 

  869. /root/dataset/class_y/sample123.png</p>
    870. 

  870. </dd>
    871. 

  871. <dt>the matching labels paths:  (That DO NOT appear in the list file)</dt><dd><p>/root/dataset/class_x/sample1.ext
    872. 

  872. /root/dataset/class_y/sample123.ext</p>
    873. 

  873. </dd>
    874. 

  874. </dl>
    875. 

  875. </div></blockquote>
    876. 

  876. </dd>
    877. 

  877. </dl>
    878. 

  878. </dd></dl>
    879. 

  879. 

  880. <dl class="py class">
    881. 

  881. <dt class="sig sig-object py" id="super_gradients.training.datasets.DirectoryDataSet">
    882. 

  882. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">DirectoryDataSet</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">root:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">samples_sub_directory:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">targets_sub_directory:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">target_extension:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">sample_loader:</span> <span class="pre">Callable</span> <span class="pre">=</span> <span class="pre">&lt;function</span> <span class="pre">default_loader&gt;</span></em>, <em class="sig-param"><span class="pre">target_loader:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">collate_fn:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">sample_extensions:</span> <span class="pre">tuple</span> <span class="pre">=</span> <span class="pre">('.jpg'</span></em>, <em class="sig-param"><span class="pre">'.jpeg'</span></em>, <em class="sig-param"><span class="pre">'.png'</span></em>, <em class="sig-param"><span class="pre">'.ppm'</span></em>, <em class="sig-param"><span class="pre">'.bmp'</span></em>, <em class="sig-param"><span class="pre">'.pgm'</span></em>, <em class="sig-param"><span class="pre">'.tif'</span></em>, <em class="sig-param"><span class="pre">'.tiff'</span></em>, <em class="sig-param"><span class="pre">'.webp')</span></em>, <em class="sig-param"><span class="pre">sample_transform:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">target_transform:</span> <span class="pre">Optional[Callable]</span> <span class="pre">=</span> <span class="pre">None</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/sg_dataset.html#DirectoryDataSet"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DirectoryDataSet" title="Permalink to this definition"></a></dt>
    883. 

  883. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    884. 

  884. <dl class="simple">
    885. 

  885. <dt>DirectoryDataSet - A PyTorch Vision Data Set extension that receives a root Dir and two separate sub directories:</dt><dd><ul class="simple">
    886. 

  886. <li><p>Sub-Directory for Samples</p></li>
    887. 

  887. <li><p>Sub-Directory for Targets</p></li>
    888. 

  888. </ul>
    889. 

  889. </dd>
    890. 

  890. </dl>
    891. 

  891. </dd></dl>
    892. 

  892. 

  893. <dl class="py class">
    894. 

  894. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationDataSet">
    895. 

  895. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">SegmentationDataSet</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">root</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">list_file</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">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">samples_sub_directory</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">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets_sub_directory</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">None</span></span></em>, <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> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">608</span></span></em>, <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">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">512</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> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">16</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">augment</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">dataset_hyper_params</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">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">cache_labels</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">cache_images</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">sample_loader</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Callable</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_loader</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Callable</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">collate_fn</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Callable</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_extension</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">'.png'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">image_mask_transforms</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torchvision.transforms.transforms.Compose</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">image_mask_transforms_aug</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torchvision.transforms.transforms.Compose</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/datasets/segmentation_datasets/segmentation_dataset.html#SegmentationDataSet"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationDataSet" title="Permalink to this definition"></a></dt>
    896. 

  896. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    897. 

  897. <dl class="py method">
    898. 

  898. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationDataSet.sample_loader">
    899. 

  899. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">sample_loader</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">sample_path</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> &#x2192; <span class="pre">&lt;module</span> <span class="pre">‘PIL.Image’</span> <span class="pre">from</span> <span class="pre">‘/Users/shaniperl/opt/anaconda3/lib/python3.9/site-packages/PIL/Image.py’&gt;</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/segmentation_dataset.html#SegmentationDataSet.sample_loader"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationDataSet.sample_loader" title="Permalink to this definition"></a></dt>
    900. 

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

  901. <dt>sample_loader - Loads a dataset image from path using PIL</dt><dd><dl class="field-list simple">
    902. 

  902. <dt class="field-odd">param sample_path</dt>
    903. 

  903. <dd class="field-odd"><p>The path to the sample image</p>
    904. 

  904. </dd>
    905. 

  905. <dt class="field-even">return</dt>
    906. 

  906. <dd class="field-even"><p>The loaded Image</p>
    907. 

  907. </dd>
    908. 

  908. </dl>
    909. 

  909. </dd>
    910. 

  910. </dl>
    911. 

  911. </dd></dl>
    912. 

  912. 

  913. <dl class="py method">
    914. 

  914. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationDataSet.sample_transform">
    915. 

  915. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">sample_transform</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">image</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/segmentation_dataset.html#SegmentationDataSet.sample_transform"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationDataSet.sample_transform" title="Permalink to this definition"></a></dt>
    916. 

  916. <dd><p>sample_transform - Transforms the sample image</p>
    917. 

  917. <blockquote>
    918. 

  918. <div><dl class="field-list simple">
    919. 

  919. <dt class="field-odd">param image</dt>
    920. 

  920. <dd class="field-odd"><p>The input image to transform</p>
    921. 

  921. </dd>
    922. 

  922. <dt class="field-even">return</dt>
    923. 

  923. <dd class="field-even"><p>The transformed image</p>
    924. 

  924. </dd>
    925. 

  925. </dl>
    926. 

  926. </div></blockquote>
    927. 

  927. </dd></dl>
    928. 

  928. 

  929. <dl class="py method">
    930. 

  930. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationDataSet.target_loader">
    931. 

  931. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">target_loader</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">target_path</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> &#x2192; <span class="pre">&lt;module</span> <span class="pre">‘PIL.Image’</span> <span class="pre">from</span> <span class="pre">‘/Users/shaniperl/opt/anaconda3/lib/python3.9/site-packages/PIL/Image.py’&gt;</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/segmentation_dataset.html#SegmentationDataSet.target_loader"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationDataSet.target_loader" title="Permalink to this definition"></a></dt>
    932. 

  932. <dd><dl class="field-list simple">
    933. 

  933. <dt class="field-odd">Parameters</dt>
    934. 

  934. <dd class="field-odd"><p><strong>target_path</strong> – The path to the sample image</p>
    935. 

  935. </dd>
    936. 

  936. <dt class="field-even">Returns</dt>
    937. 

  937. <dd class="field-even"><p>The loaded Image</p>
    938. 

  938. </dd>
    939. 

  939. </dl>
    940. 

  940. </dd></dl>
    941. 

  941. 

  942. <dl class="py method">
    943. 

  943. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationDataSet.target_transform">
    944. 

  944. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">target_transform</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">target</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/segmentation_dataset.html#SegmentationDataSet.target_transform"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationDataSet.target_transform" title="Permalink to this definition"></a></dt>
    945. 

  945. <dd><p>target_transform - Transforms the sample image</p>
    946. 

  946. <blockquote>
    947. 

  947. <div><dl class="field-list simple">
    948. 

  948. <dt class="field-odd">param target</dt>
    949. 

  949. <dd class="field-odd"><p>The target mask to transform</p>
    950. 

  950. </dd>
    951. 

  951. <dt class="field-even">return</dt>
    952. 

  952. <dd class="field-even"><p>The transformed target mask</p>
    953. 

  953. </dd>
    954. 

  954. </dl>
    955. 

  955. </div></blockquote>
    956. 

  956. </dd></dl>
    957. 

  957. 

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

  959. 

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

  961. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalVOC2012SegmentationDataSet">
    962. 

  962. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">PascalVOC2012SegmentationDataSet</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">sample_suffix</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_suffix</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">args</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/datasets/segmentation_datasets/pascal_voc_segmentation.html#PascalVOC2012SegmentationDataSet"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalVOC2012SegmentationDataSet" title="Permalink to this definition"></a></dt>
    963. 

  963. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    964. 

  964. <p>PascalVOC2012SegmentationDataSet - Segmentation Data Set Class for Pascal VOC 2012 Data Set</p>
    965. 

  965. <dl class="py method">
    966. 

  966. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalVOC2012SegmentationDataSet.decode_segmentation_mask">
    967. 

  967. <span class="sig-name descname"><span class="pre">decode_segmentation_mask</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">label_mask</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">numpy.ndarray</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/pascal_voc_segmentation.html#PascalVOC2012SegmentationDataSet.decode_segmentation_mask"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalVOC2012SegmentationDataSet.decode_segmentation_mask" title="Permalink to this definition"></a></dt>
    968. 

  968. <dd><dl class="simple">
    969. 

  969. <dt>decode_segmentation_mask - Decodes the colors for the Segmentation Mask</dt><dd><dl class="field-list simple">
    970. 

  970. <dt class="field-odd">param</dt>
    971. 

  971. <dd class="field-odd"><p>label_mask:  an (M,N) array of integer values denoting
    972. 

  972. the class label at each spatial location.</p>
    973. 

  973. </dd>
    974. 

  974. </dl>
    975. 

  975. </dd>
    976. 

  976. </dl>
    977. 

  977. <dl class="field-list simple">
    978. 

  978. <dt class="field-odd">Returns</dt>
    979. 

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

  980. </dd>
    981. 

  981. </dl>
    982. 

  982. </dd></dl>
    983. 

  983. 

  984. </dd></dl>
    985. 

  985. 

  986. <dl class="py class">
    987. 

  987. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalAUG2012SegmentationDataSet">
    988. 

  988. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">PascalAUG2012SegmentationDataSet</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">args</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/datasets/segmentation_datasets/pascal_aug_segmentation.html#PascalAUG2012SegmentationDataSet"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalAUG2012SegmentationDataSet" title="Permalink to this definition"></a></dt>
    989. 

  989. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    990. 

  990. <p>PascalAUG2012SegmentationDataSet - Segmentation Data Set Class for Pascal AUG 2012 Data Set</p>
    991. 

  991. <dl class="py method">
    992. 

  992. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalAUG2012SegmentationDataSet.target_loader">
    993. 

  993. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">target_loader</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">target_path</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> &#x2192; <span class="pre">&lt;module</span> <span class="pre">‘PIL.Image’</span> <span class="pre">from</span> <span class="pre">‘/Users/shaniperl/opt/anaconda3/lib/python3.9/site-packages/PIL/Image.py’&gt;</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/pascal_aug_segmentation.html#PascalAUG2012SegmentationDataSet.target_loader"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalAUG2012SegmentationDataSet.target_loader" title="Permalink to this definition"></a></dt>
    994. 

  994. <dd><dl class="field-list simple">
    995. 

  995. <dt class="field-odd">Parameters</dt>
    996. 

  996. <dd class="field-odd"><p><strong>target_path</strong> – The path to the target data</p>
    997. 

  997. </dd>
    998. 

  998. <dt class="field-even">Returns</dt>
    999. 

  999. <dd class="field-even"><p>The loaded target</p>
    1000. 

  1000. </dd>
    1001. 

  1001. </dl>
    1002. 

  1002. </dd></dl>
    1003. 

  1003. 

  1004. </dd></dl>
    1005. 

  1005. 

  1006. <dl class="py class">
    1007. 

  1007. <dt class="sig sig-object py" id="super_gradients.training.datasets.CoCoSegmentationDataSet">
    1008. 

  1008. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">CoCoSegmentationDataSet</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_classes_inclusion_tuples_list</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">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">None</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</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/datasets/segmentation_datasets/coco_segmentation.html#CoCoSegmentationDataSet"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.CoCoSegmentationDataSet" title="Permalink to this definition"></a></dt>
    1009. 

  1009. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    1010. 

  1010. <p>CoCoSegmentationDataSet - Segmentation Data Set Class for COCO 2017 Segmentation Data Set</p>
    1011. 

  1011. <dl class="py method">
    1012. 

  1012. <dt class="sig sig-object py" id="super_gradients.training.datasets.CoCoSegmentationDataSet.target_loader">
    1013. 

  1013. <span class="sig-name descname"><span class="pre">target_loader</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">mask_metadata_tuple</span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">&lt;module</span> <span class="pre">‘PIL.Image’</span> <span class="pre">from</span> <span class="pre">‘/Users/shaniperl/opt/anaconda3/lib/python3.9/site-packages/PIL/Image.py’&gt;</span><a class="reference internal" href="_modules/super_gradients/training/datasets/segmentation_datasets/coco_segmentation.html#CoCoSegmentationDataSet.target_loader"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.CoCoSegmentationDataSet.target_loader" title="Permalink to this definition"></a></dt>
    1014. 

  1014. <dd><dl class="field-list simple">
    1015. 

  1015. <dt class="field-odd">Parameters</dt>
    1016. 

  1016. <dd class="field-odd"><p><strong>mask_metadata_tuple</strong> – A tuple of (coco_image_id, original_image_height, original_image_width)</p>
    1017. 

  1017. </dd>
    1018. 

  1018. <dt class="field-even">Returns</dt>
    1019. 

  1019. <dd class="field-even"><p>The mask image created from the array</p>
    1020. 

  1020. </dd>
    1021. 

  1021. </dl>
    1022. 

  1022. </dd></dl>
    1023. 

  1023. 

  1024. </dd></dl>
    1025. 

  1025. 

  1026. <dl class="py class">
    1027. 

  1027. <dt class="sig sig-object py" id="super_gradients.training.datasets.TestDatasetInterface">
    1028. 

  1028. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">TestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">trainset</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dataset_params</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">classes</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/datasets/dataset_interfaces/dataset_interface.html#TestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.TestDatasetInterface" title="Permalink to this definition"></a></dt>
    1029. 

  1029. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    1030. 

  1030. <dl class="py method">
    1031. 

  1031. <dt class="sig sig-object py" id="super_gradients.training.datasets.TestDatasetInterface.get_data_loaders">
    1032. 

  1032. <span class="sig-name descname"><span class="pre">get_data_loaders</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">batch_size_factor</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">num_workers</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">8</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">train_batch_size</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">val_batch_size</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">distributed_sampler</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/datasets/dataset_interfaces/dataset_interface.html#TestDatasetInterface.get_data_loaders"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.TestDatasetInterface.get_data_loaders" title="Permalink to this definition"></a></dt>
    1033. 

  1033. <dd><p>Get self.train_loader, self.val_loader, self.test_loader, self.classes.</p>
    1034. 

  1034. <p>If the data loaders haven’t been initialized yet, build them first.</p>
    1035. 

  1035. <dl class="field-list simple">
    1036. 

  1036. <dt class="field-odd">Parameters</dt>
    1037. 

  1037. <dd class="field-odd"><p><strong>kwargs</strong> – kwargs are passed to build_data_loaders.</p>
    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.datasets.DatasetInterface">
    1046. 

  1046. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">DatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">train_loader</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">val_loader</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">test_loader</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">classes</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/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface" title="Permalink to this definition"></a></dt>
    1047. 

  1047. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    1048. 

  1048. <p>DatasetInterface - This class manages all of the “communiation” the Model has with the Data Sets</p>
    1049. 

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

  1050. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.download_from_cloud">
    1051. 

  1051. <span class="sig-name descname"><span class="pre">download_from_cloud</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.download_from_cloud"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.download_from_cloud" title="Permalink to this definition"></a></dt>
    1052. 

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

  1053. 

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

  1055. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.build_data_loaders">
    1056. 

  1056. <span class="sig-name descname"><span class="pre">build_data_loaders</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">batch_size_factor</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">num_workers</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">8</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">train_batch_size</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">val_batch_size</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">test_batch_size</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">distributed_sampler</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/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.build_data_loaders"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.build_data_loaders" title="Permalink to this definition"></a></dt>
    1057. 

  1057. <dd><p>define train, val (and optionally test) loaders. The method deals separately with distributed training and standard
    1058. 

  1058. (non distributed, or parallel training). In the case of distributed training we need to rely on distributed
    1059. 

  1059. samplers.
    1060. 

  1060. :param batch_size_factor: int - factor to multiply the batch size (usually for multi gpu)
    1061. 

  1061. :param num_workers: int - number of workers (parallel processes) for dataloaders
    1062. 

  1062. :param train_batch_size: int - batch size for train loader, if None will be taken from dataset_params
    1063. 

  1063. :param val_batch_size: int - batch size for val loader, if None will be taken from dataset_params
    1064. 

  1064. :param distributed_sampler: boolean flag for distributed training mode
    1065. 

  1065. :return: train_loader, val_loader, classes: list of classes</p>
    1066. 

  1066. </dd></dl>
    1067. 

  1067. 

  1068. <dl class="py method">
    1069. 

  1069. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.get_data_loaders">
    1070. 

  1070. <span class="sig-name descname"><span class="pre">get_data_loaders</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/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.get_data_loaders"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.get_data_loaders" title="Permalink to this definition"></a></dt>
    1071. 

  1071. <dd><p>Get self.train_loader, self.val_loader, self.test_loader, self.classes.</p>
    1072. 

  1072. <p>If the data loaders haven’t been initialized yet, build them first.</p>
    1073. 

  1073. <dl class="field-list simple">
    1074. 

  1074. <dt class="field-odd">Parameters</dt>
    1075. 

  1075. <dd class="field-odd"><p><strong>kwargs</strong> – kwargs are passed to build_data_loaders.</p>
    1076. 

  1076. </dd>
    1077. 

  1077. </dl>
    1078. 

  1078. </dd></dl>
    1079. 

  1079. 

  1080. <dl class="py method">
    1081. 

  1081. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.get_val_sample">
    1082. 

  1082. <span class="sig-name descname"><span class="pre">get_val_sample</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">num_samples</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/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.get_val_sample"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.get_val_sample" title="Permalink to this definition"></a></dt>
    1083. 

  1083. <dd></dd></dl>
    1084. 

  1084. 

  1085. <dl class="py method">
    1086. 

  1086. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.get_dataset_params">
    1087. 

  1087. <span class="sig-name descname"><span class="pre">get_dataset_params</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.get_dataset_params"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.get_dataset_params" title="Permalink to this definition"></a></dt>
    1088. 

  1088. <dd></dd></dl>
    1089. 

  1089. 

  1090. <dl class="py method">
    1091. 

  1091. <dt class="sig sig-object py" id="super_gradients.training.datasets.DatasetInterface.print_dataset_details">
    1092. 

  1092. <span class="sig-name descname"><span class="pre">print_dataset_details</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#DatasetInterface.print_dataset_details"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DatasetInterface.print_dataset_details" title="Permalink to this definition"></a></dt>
    1093. 

  1093. <dd></dd></dl>
    1094. 

  1094. 

  1095. </dd></dl>
    1096. 

  1096. 

  1097. <dl class="py class">
    1098. 

  1098. <dt class="sig sig-object py" id="super_gradients.training.datasets.Cifar10DatasetInterface">
    1099. 

  1099. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">Cifar10DatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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/datasets/dataset_interfaces/dataset_interface.html#Cifar10DatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.Cifar10DatasetInterface" title="Permalink to this definition"></a></dt>
    1100. 

  1100. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.LibraryDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.LibraryDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.LibraryDatasetInterface</span></code></a></p>
    1101. 

  1101. </dd></dl>
    1102. 

  1102. 

  1103. <dl class="py class">
    1104. 

  1104. <dt class="sig sig-object py" id="super_gradients.training.datasets.CoCoSegmentationDatasetInterface">
    1105. 

  1105. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">CoCoSegmentationDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">cache_labels</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">cache_images</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">dataset_classes_inclusion_tuples_list</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">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">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#CoCoSegmentationDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.CoCoSegmentationDatasetInterface" title="Permalink to this definition"></a></dt>
    1106. 

  1106. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.CoCoDataSetInterfaceBase" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.CoCoDataSetInterfaceBase"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.CoCoDataSetInterfaceBase</span></code></a></p>
    1107. 

  1107. </dd></dl>
    1108. 

  1108. 

  1109. <dl class="py class">
    1110. 

  1110. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalVOC2012SegmentationDataSetInterface">
    1111. 

  1111. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">PascalVOC2012SegmentationDataSetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">cache_labels</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">cache_images</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/datasets/dataset_interfaces/dataset_interface.html#PascalVOC2012SegmentationDataSetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalVOC2012SegmentationDataSetInterface" title="Permalink to this definition"></a></dt>
    1112. 

  1112. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    1113. 

  1113. </dd></dl>
    1114. 

  1114. 

  1115. <dl class="py class">
    1116. 

  1116. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalAUG2012SegmentationDataSetInterface">
    1117. 

  1117. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">PascalAUG2012SegmentationDataSetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">cache_labels</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">cache_images</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/datasets/dataset_interfaces/dataset_interface.html#PascalAUG2012SegmentationDataSetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalAUG2012SegmentationDataSetInterface" title="Permalink to this definition"></a></dt>
    1118. 

  1118. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    1119. 

  1119. </dd></dl>
    1120. 

  1120. 

  1121. <dl class="py class">
    1122. 

  1122. <dt class="sig sig-object py" id="super_gradients.training.datasets.TestYoloDetectionDatasetInterface">
    1123. 

  1123. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">TestYoloDetectionDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">input_dims</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">(3,</span> <span class="pre">32,</span> <span class="pre">32)</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">5</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#TestYoloDetectionDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.TestYoloDetectionDatasetInterface" title="Permalink to this definition"></a></dt>
    1124. 

  1124. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    1125. 

  1125. <p>note: the output size is (batch_size, 6) in the test while in real training
    1126. 

  1126. the size of axis 0 can vary (the number of bounding boxes)</p>
    1127. 

  1127. </dd></dl>
    1128. 

  1128. 

  1129. <dl class="py class">
    1130. 

  1130. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionTestDatasetInterface">
    1131. 

  1131. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">DetectionTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">320</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">4</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#DetectionTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionTestDatasetInterface" title="Permalink to this definition"></a></dt>
    1132. 

  1132. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    1133. 

  1133. </dd></dl>
    1134. 

  1134. 

  1135. <dl class="py class">
    1136. 

  1136. <dt class="sig sig-object py" id="super_gradients.training.datasets.ClassificationTestDatasetInterface">
    1137. 

  1137. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">ClassificationTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">32</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">5</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#ClassificationTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.ClassificationTestDatasetInterface" title="Permalink to this definition"></a></dt>
    1138. 

  1138. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    1139. 

  1139. </dd></dl>
    1140. 

  1140. 

  1141. <dl class="py class">
    1142. 

  1142. <dt class="sig sig-object py" id="super_gradients.training.datasets.SegmentationTestDatasetInterface">
    1143. 

  1143. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">SegmentationTestDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">image_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">512</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">4</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#SegmentationTestDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.SegmentationTestDatasetInterface" title="Permalink to this definition"></a></dt>
    1144. 

  1144. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.TestDatasetInterface</span></code></a></p>
    1145. 

  1145. </dd></dl>
    1146. 

  1146. 

  1147. <dl class="py class">
    1148. 

  1148. <dt class="sig sig-object py" id="super_gradients.training.datasets.ImageNetDatasetInterface">
    1149. 

  1149. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">ImageNetDatasetInterface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_params</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">data_dir</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'/data/Imagenet'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/dataset_interfaces/dataset_interface.html#ImageNetDatasetInterface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.ImageNetDatasetInterface" title="Permalink to this definition"></a></dt>
    1150. 

  1150. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></code></a></p>
    1151. 

  1151. </dd></dl>
    1152. 

  1152. 

  1153. <dl class="py class">
    1154. 

  1154. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset">
    1155. 

  1155. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">DetectionDataset</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data_dir</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">input_dim</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">tuple</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">original_target_format</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.utils.html#super_gradients.training.utils.detection_utils.DetectionTargetsFormat" title="super_gradients.training.utils.detection_utils.DetectionTargetsFormat"><span class="pre">super_gradients.training.utils.detection_utils.DetectionTargetsFormat</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_num_samples</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">cache</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">cache_path</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">str</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">transforms</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">super_gradients.training.transforms.transforms.DetectionTransform</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">all_classes_list</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">List</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">class_inclusion_list</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">List</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_empty_annotations</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">target_fields</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">List</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">output_fields</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">List</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset" title="Permalink to this definition"></a></dt>
    1156. 

  1156. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    1157. 

  1157. <p>Detection dataset.</p>
    1158. 

  1158. <p>This is a boilerplate class to facilitate the implementation of datasets.</p>
    1159. 

  1159. <dl>
    1160. 

  1160. <dt>HOW TO CREATE A DATASET THAT INHERITS FROM DetectionDataSet ?</dt><dd><ul class="simple">
    1161. 

  1161. <li><p>Inherit from DetectionDataSet</p></li>
    1162. 

  1162. <li><p>implement the method self._load_annotation to return at least the fields “target” and “img_path”</p></li>
    1163. 

  1163. <li><dl class="simple">
    1164. 

  1164. <dt>Call super().__init__ with the required params.</dt><dd><dl class="simple">
    1165. 

  1165. <dt>//!super().__init__ will call self._load_annotation, so make sure that every required</dt><dd><p>attributes are set up before calling super().__init__ (ideally just call it last)</p>
    1166. 

  1166. </dd>
    1167. 

  1167. </dl>
    1168. 

  1168. </dd>
    1169. 

  1169. </dl>
    1170. 

  1170. </li>
    1171. 

  1171. </ul>
    1172. 

  1172. </dd>
    1173. 

  1173. <dt>WORKFLOW:</dt><dd><ul class="simple">
    1174. 

  1174. <li><dl class="simple">
    1175. 

  1175. <dt>On instantiation:</dt><dd><ul>
    1176. 

  1176. <li><p>All annotations are cached. If class_inclusion_list was specified, there is also subclassing at this step.</p></li>
    1177. 

  1177. <li><p>If cache is True, the images are also cached</p></li>
    1178. 

  1178. </ul>
    1179. 

  1179. </dd>
    1180. 

  1180. </dl>
    1181. 

  1181. </li>
    1182. 

  1182. <li><dl class="simple">
    1183. 

  1183. <dt>On call (__getitem__) for a specific image index:</dt><dd><ul>
    1184. 

  1184. <li><p>The image and annotations are grouped together in a dict called SAMPLE</p></li>
    1185. 

  1185. <li><p>the sample is processed according to th transform</p></li>
    1186. 

  1186. <li><p>Only the specified fields are returned by __getitem__</p></li>
    1187. 

  1187. </ul>
    1188. 

  1188. </dd>
    1189. 

  1189. </dl>
    1190. 

  1190. </li>
    1191. 

  1191. </ul>
    1192. 

  1192. </dd>
    1193. 

  1193. <dt>TERMINOLOGY</dt><dd><ul>
    1194. 

  1194. <li><p>TARGET:       Groundtruth, made of bboxes. The format can vary from one dataset to another</p></li>
    1195. 

  1195. <li><dl class="simple">
    1196. 

  1196. <dt>ANNOTATION:   Combination of targets (groundtruth) and metadata of the image, but without the image itself.</dt><dd><p>&gt; Has to include the fields “target” and “img_path”
    1197. 

  1197. &gt; Can include other fields like “crowd_target”, “image_info”, “segmentation”, …</p>
    1198. 

  1198. </dd>
    1199. 

  1199. </dl>
    1200. 

  1200. </li>
    1201. 

  1201. <li><dl class="simple">
    1202. 

  1202. <dt>SAMPLE:       Outout of the dataset:</dt><dd><p>&gt; Has to include the fields “target” and “image”
    1203. 

  1203. &gt; Can include other fields like “crowd_target”, “image_info”, “segmentation”, …</p>
    1204. 

  1204. </dd>
    1205. 

  1205. </dl>
    1206. 

  1206. </li>
    1207. 

  1207. <li><p>INDEX:        Refers to the index in the dataset.</p></li>
    1208. 

  1208. <li><dl>
    1209. 

  1209. <dt>SAMPLE ID:    Refers to the id of sample before droping any annotaion.</dt><dd><p>Let’s imagine a situation where the downloaded data is made of 120 images but 20 were drop
    1210. 

  1210. because they had no annotation. In that case:</p>
    1211. 

  1211. <blockquote>
    1212. 

  1212. <div><p>&gt; We have 120 samples so sample_id will be between 0 and 119
    1213. 

  1213. &gt; But only 100 will be indexed so index will be between 0 and 99
    1214. 

  1214. &gt; Therefore, we also have len(self) = 100</p>
    1215. 

  1215. </div></blockquote>
    1216. 

  1216. </dd>
    1217. 

  1217. </dl>
    1218. 

  1218. </li>
    1219. 

  1219. </ul>
    1220. 

  1220. </dd>
    1221. 

  1221. </dl>
    1222. 

  1222. <dl class="py method">
    1223. 

  1223. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.get_random_item">
    1224. 

  1224. <span class="sig-name descname"><span class="pre">get_random_item</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset.get_random_item"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.get_random_item" title="Permalink to this definition"></a></dt>
    1225. 

  1225. <dd></dd></dl>
    1226. 

  1226. 

  1227. <dl class="py method">
    1228. 

  1228. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.get_sample">
    1229. 

  1229. <span class="sig-name descname"><span class="pre">get_sample</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">index</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> &#x2192; <span class="pre">Dict</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">numpy.ndarray</span><span class="p"><span class="pre">,</span> </span><span class="pre">Any</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset.get_sample"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.get_sample" title="Permalink to this definition"></a></dt>
    1230. 

  1230. <dd><p>Get raw sample, before any transform (beside subclassing).
    1231. 

  1231. :param index:   Image index
    1232. 

  1232. :return:        Sample, i.e. a dictionary including at least “image” and “target”</p>
    1233. 

  1233. </dd></dl>
    1234. 

  1234. 

  1235. <dl class="py method">
    1236. 

  1236. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.get_resized_image">
    1237. 

  1237. <span class="sig-name descname"><span class="pre">get_resized_image</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">index</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> &#x2192; <span class="pre">numpy.ndarray</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset.get_resized_image"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.get_resized_image" title="Permalink to this definition"></a></dt>
    1238. 

  1238. <dd><p>Get the resized image at a specific sample_id, either from cache or by loading from disk, based on self.cached_imgs
    1239. 

  1239. :param index:  Image index
    1240. 

  1240. :return:       Resized image</p>
    1241. 

  1241. </dd></dl>
    1242. 

  1242. 

  1243. <dl class="py method">
    1244. 

  1244. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.apply_transforms">
    1245. 

  1245. <span class="sig-name descname"><span class="pre">apply_transforms</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">sample</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Dict</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">numpy.ndarray</span><span class="p"><span class="pre">,</span> </span><span class="pre">Any</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">Dict</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">numpy.ndarray</span><span class="p"><span class="pre">,</span> </span><span class="pre">Any</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset.apply_transforms"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.apply_transforms" title="Permalink to this definition"></a></dt>
    1246. 

  1246. <dd><p>Applies self.transforms sequentially to sample</p>
    1247. 

  1247. <dl class="simple">
    1248. 

  1248. <dt>If a transforms has the attribute ‘additional_samples_count’, additional samples will be loaded and stored in</dt><dd><p>sample[“additional_samples”] prior to applying it. Combining with the attribute “non_empty_annotations” will load
    1249. 

  1249. only additional samples with objects in them.</p>
    1250. 

  1250. </dd>
    1251. 

  1251. </dl>
    1252. 

  1252. <dl class="field-list simple">
    1253. 

  1253. <dt class="field-odd">Parameters</dt>
    1254. 

  1254. <dd class="field-odd"><p><strong>sample</strong> – Sample to apply the transforms on to (loaded with self.get_sample)</p>
    1255. 

  1255. </dd>
    1256. 

  1256. <dt class="field-even">Returns</dt>
    1257. 

  1257. <dd class="field-even"><p>Transformed sample</p>
    1258. 

  1258. </dd>
    1259. 

  1259. </dl>
    1260. 

  1260. </dd></dl>
    1261. 

  1261. 

  1262. <dl class="py method">
    1263. 

  1263. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.get_random_samples">
    1264. 

  1264. <span class="sig-name descname"><span class="pre">get_random_samples</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">count</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">non_empty_annotations_only</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> &#x2192; <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 class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">numpy.ndarray</span><span class="p"><span class="pre">,</span> </span><span class="pre">Any</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/detection_dataset.html#DetectionDataset.get_random_samples"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.get_random_samples" title="Permalink to this definition"></a></dt>
    1265. 

  1265. <dd><p>Load random samples.</p>
    1266. 

  1266. <dl class="field-list simple">
    1267. 

  1267. <dt class="field-odd">Parameters</dt>
    1268. 

  1268. <dd class="field-odd"><ul class="simple">
    1269. 

  1269. <li><p><strong>count</strong> – The number of samples wanted</p></li>
    1270. 

  1270. <li><p><strong>non_empty_annotations_only</strong> – If true, only return samples with at least 1 annotation</p></li>
    1271. 

  1271. </ul>
    1272. 

  1272. </dd>
    1273. 

  1273. <dt class="field-even">Returns</dt>
    1274. 

  1274. <dd class="field-even"><p>A list of samples satisfying input params</p>
    1275. 

  1275. </dd>
    1276. 

  1276. </dl>
    1277. 

  1277. </dd></dl>
    1278. 

  1278. 

  1279. <dl class="py method">
    1280. 

  1280. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.get_random_sample">
    1281. 

  1281. <span class="sig-name descname"><span class="pre">get_random_sample</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">non_empty_annotations_only</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/datasets/detection_datasets/detection_dataset.html#DetectionDataset.get_random_sample"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.get_random_sample" title="Permalink to this definition"></a></dt>
    1282. 

  1282. <dd></dd></dl>
    1283. 

  1283. 

  1284. <dl class="py property">
    1285. 

  1285. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.output_target_format">
    1286. 

  1286. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">output_target_format</span></span><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.output_target_format" title="Permalink to this definition"></a></dt>
    1287. 

  1287. <dd></dd></dl>
    1288. 

  1288. 

  1289. <dl class="py method">
    1290. 

  1290. <dt class="sig sig-object py" id="super_gradients.training.datasets.DetectionDataset.plot">
    1291. 

  1291. <span class="sig-name descname"><span class="pre">plot</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_samples_per_plot</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">16</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">n_plots</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">plot_transformed_data</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/datasets/detection_datasets/detection_dataset.html#DetectionDataset.plot"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.DetectionDataset.plot" title="Permalink to this definition"></a></dt>
    1292. 

  1292. <dd><p>Combine samples of images with bbox into plots and display the result.</p>
    1293. 

  1293. <dl class="field-list simple">
    1294. 

  1294. <dt class="field-odd">Parameters</dt>
    1295. 

  1295. <dd class="field-odd"><ul class="simple">
    1296. 

  1296. <li><p><strong>max_samples_per_plot</strong> – Maximum number of images to be displayed per plot</p></li>
    1297. 

  1297. <li><p><strong>n_plots</strong> – Number of plots to display (each plot being a combination of img with bbox)</p></li>
    1298. 

  1298. <li><p><strong>plot_transformed_data</strong> – If True, the plot will be over samples after applying transforms (i.e. on __getitem__).
    1299. 

  1299. If False, the plot will be over the raw samples (i.e. on get_sample)</p></li>
    1300. 

  1300. </ul>
    1301. 

  1301. </dd>
    1302. 

  1302. <dt class="field-even">Returns</dt>
    1303. 

  1303. <dd class="field-even"><p></p>
    1304. 

  1304. </dd>
    1305. 

  1305. </dl>
    1306. 

  1306. </dd></dl>
    1307. 

  1307. 

  1308. </dd></dl>
    1309. 

  1309. 

  1310. <dl class="py class">
    1311. 

  1311. <dt class="sig sig-object py" id="super_gradients.training.datasets.COCODetectionDataset">
    1312. 

  1312. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">COCODetectionDataset</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">tuple</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">data_dir</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">str</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">json_file</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">'instances_train2017.json'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">name</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">'images/train2017'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cache</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">cache_dir_path</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">str</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">tight_box_rotation</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">transforms</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>, <em class="sig-param"><span class="n"><span class="pre">with_crowd</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/datasets/detection_datasets/coco_detection.html#COCODetectionDataset"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.COCODetectionDataset" title="Permalink to this definition"></a></dt>
    1313. 

  1313. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    1314. 

  1314. <p>Detection dataset COCO implementation</p>
    1315. 

  1315. <dl class="py method">
    1316. 

  1316. <dt class="sig sig-object py" id="super_gradients.training.datasets.COCODetectionDataset.load_resized_img">
    1317. 

  1317. <span class="sig-name descname"><span class="pre">load_resized_img</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">index</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/coco_detection.html#COCODetectionDataset.load_resized_img"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.COCODetectionDataset.load_resized_img" title="Permalink to this definition"></a></dt>
    1318. 

  1318. <dd><p>Loads image at index, and resizes it to self.input_dim</p>
    1319. 

  1319. <dl class="field-list simple">
    1320. 

  1320. <dt class="field-odd">Parameters</dt>
    1321. 

  1321. <dd class="field-odd"><p><strong>index</strong> – index to load the image from</p>
    1322. 

  1322. </dd>
    1323. 

  1323. <dt class="field-even">Returns</dt>
    1324. 

  1324. <dd class="field-even"><p>resized_img</p>
    1325. 

  1325. </dd>
    1326. 

  1326. </dl>
    1327. 

  1327. </dd></dl>
    1328. 

  1328. 

  1329. <dl class="py method">
    1330. 

  1330. <dt class="sig sig-object py" id="super_gradients.training.datasets.COCODetectionDataset.load_sample">
    1331. 

  1331. <span class="sig-name descname"><span class="pre">load_sample</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">index</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/coco_detection.html#COCODetectionDataset.load_sample"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.COCODetectionDataset.load_sample" title="Permalink to this definition"></a></dt>
    1332. 

  1332. <dd><dl>
    1333. 

  1333. <dt>Loads sample at self.ids[index] as dictionary that holds:</dt><dd><p>“image”: Image resized to self.input_dim
    1334. 

  1334. “target”: Detection ground truth, np.array shaped (num_targets, 5), format is [class,x1,y1,x2,y2] with</p>
    1335. 

  1335. <blockquote>
    1336. 

  1336. <div><p>image coordinates.</p>
    1337. 

  1337. </div></blockquote>
    1338. 

  1338. <p>“target_seg”: Segmentation map convex hull derived detection target.
    1339. 

  1339. “info”: Original shape (height,width).
    1340. 

  1340. “id”: COCO image id</p>
    1341. 

  1341. </dd>
    1342. 

  1342. </dl>
    1343. 

  1343. <dl class="field-list simple">
    1344. 

  1344. <dt class="field-odd">Parameters</dt>
    1345. 

  1345. <dd class="field-odd"><p><strong>index</strong> – Sample index</p>
    1346. 

  1346. </dd>
    1347. 

  1347. <dt class="field-even">Returns</dt>
    1348. 

  1348. <dd class="field-even"><p>sample as described above</p>
    1349. 

  1349. </dd>
    1350. 

  1350. </dl>
    1351. 

  1351. </dd></dl>
    1352. 

  1352. 

  1353. <dl class="py method">
    1354. 

  1354. <dt class="sig sig-object py" id="super_gradients.training.datasets.COCODetectionDataset.load_image">
    1355. 

  1355. <span class="sig-name descname"><span class="pre">load_image</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">index</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/datasets/detection_datasets/coco_detection.html#COCODetectionDataset.load_image"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.COCODetectionDataset.load_image" title="Permalink to this definition"></a></dt>
    1356. 

  1356. <dd><p>Loads image at index with its original resolution
    1357. 

  1357. :param index: index in self.annotations
    1358. 

  1358. :return: image (np.array)</p>
    1359. 

  1359. </dd></dl>
    1360. 

  1360. 

  1361. <dl class="py method">
    1362. 

  1362. <dt class="sig sig-object py" id="super_gradients.training.datasets.COCODetectionDataset.apply_transforms">
    1363. 

  1363. <span class="sig-name descname"><span class="pre">apply_transforms</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">sample</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/datasets/detection_datasets/coco_detection.html#COCODetectionDataset.apply_transforms"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.COCODetectionDataset.apply_transforms" title="Permalink to this definition"></a></dt>
    1364. 

  1364. <dd><p>Applies self.transforms sequentially to sample</p>
    1365. 

  1365. <dl class="simple">
    1366. 

  1366. <dt>If a transforms has the attribute ‘additional_samples_count’, additional samples will be loaded and stored in</dt><dd><p>sample[“additional_samples”] prior to applying it. Combining with the attribute “non_empty_targets” will load
    1367. 

  1367. only additional samples with objects in them.</p>
    1368. 

  1368. </dd>
    1369. 

  1369. </dl>
    1370. 

  1370. <dl class="field-list simple">
    1371. 

  1371. <dt class="field-odd">Parameters</dt>
    1372. 

  1372. <dd class="field-odd"><p><strong>sample</strong> – Sample to apply the transforms on to (loaded with self.load_sample)</p>
    1373. 

  1373. </dd>
    1374. 

  1374. <dt class="field-even">Returns</dt>
    1375. 

  1375. <dd class="field-even"><p>Transformed sample</p>
    1376. 

  1376. </dd>
    1377. 

  1377. </dl>
    1378. 

  1378. </dd></dl>
    1379. 

  1379. 

  1380. </dd></dl>
    1381. 

  1381. 

  1382. <dl class="py class">
    1383. 

  1383. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalVOCDetectionDataset">
    1384. 

  1384. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.datasets.</span></span><span class="sig-name descname"><span class="pre">PascalVOCDetectionDataset</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">images_sub_directory</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="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</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/datasets/detection_datasets/pascal_voc_detection.html#PascalVOCDetectionDataset"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalVOCDetectionDataset" title="Permalink to this definition"></a></dt>
    1385. 

  1385. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">Generic</span></code>[<code class="xref py py-obj docutils literal notranslate"><span class="pre">torch.utils.data.dataset.T_co</span></code>]</p>
    1386. 

  1386. <p>Dataset for Pascal VOC object detection</p>
    1387. 

  1387. <dl class="py method">
    1388. 

  1388. <dt class="sig sig-object py" id="super_gradients.training.datasets.PascalVOCDetectionDataset.download">
    1389. 

  1389. <em class="property"><span class="pre">static</span> </em><span class="sig-name descname"><span class="pre">download</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data_dir</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/datasets/detection_datasets/pascal_voc_detection.html#PascalVOCDetectionDataset.download"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.datasets.PascalVOCDetectionDataset.download" title="Permalink to this definition"></a></dt>
    1390. 

  1390. <dd><p>Download Pascal dataset in XYXY_LABEL format.</p>
    1391. 

  1391. <p>Data extracted form <a class="reference external" href="http://host.robots.ox.ac.uk/pascal/VOC/">http://host.robots.ox.ac.uk/pascal/VOC/</a></p>
    1392. 

  1392. </dd></dl>
    1393. 

  1393. 

  1394. </dd></dl>
    1395. 

  1395. 

  1396. </section>
    1397. 

  1397. <section id="super-gradients-training-exceptions-module">
    1398. 

  1398. <h2>super_gradients.training.exceptions module<a class="headerlink" href="#super-gradients-training-exceptions-module" title="Permalink to this headline"></a></h2>
    1399. 

  1399. <span class="target" id="module-super_gradients.training.exceptions"></span></section>
    1400. 

  1400. <section id="module-super_gradients.training.legacy">
    1401. 

  1401. <span id="super-gradients-training-legacy-module"></span><h2>super_gradients.training.legacy module<a class="headerlink" href="#module-super_gradients.training.legacy" title="Permalink to this headline"></a></h2>
    1402. 

  1402. </section>
    1403. 

  1403. <section id="module-super_gradients.training.losses">
    1404. 

  1404. <span id="super-gradients-training-losses-models-module"></span><h2>super_gradients.training.losses_models module<a class="headerlink" href="#module-super_gradients.training.losses" title="Permalink to this headline"></a></h2>
    1405. 

  1405. <dl class="py class">
    1406. 

  1406. <dt class="sig sig-object py" id="super_gradients.training.losses.FocalLoss">
    1407. 

  1407. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">FocalLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">loss_fcn</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.loss.BCEWithLogitsLoss</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gamma</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.25</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/focal_loss.html#FocalLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.FocalLoss" title="Permalink to this definition"></a></dt>
    1408. 

  1408. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1409. 

  1409. <p>Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)</p>
    1410. 

  1410. <dl class="py attribute">
    1411. 

  1411. <dt class="sig sig-object py" id="super_gradients.training.losses.FocalLoss.reduction">
    1412. 

  1412. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.FocalLoss.reduction" title="Permalink to this definition"></a></dt>
    1413. 

  1413. <dd></dd></dl>
    1414. 

  1414. 

  1415. <dl class="py method">
    1416. 

  1416. <dt class="sig sig-object py" id="super_gradients.training.losses.FocalLoss.forward">
    1417. 

  1417. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">pred</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">true</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/focal_loss.html#FocalLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.FocalLoss.forward" title="Permalink to this definition"></a></dt>
    1418. 

  1418. <dd><p>Defines the computation performed at every call.</p>
    1419. 

  1419. <p>Should be overridden by all subclasses.</p>
    1420. 

  1420. <div class="admonition note">
    1421. 

  1421. <p class="admonition-title">Note</p>
    1422. 

  1422. <p>Although the recipe for forward pass needs to be defined within
    1423. 

  1423. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1424. 

  1424. instead of this since the former takes care of running the
    1425. 

  1425. registered hooks while the latter silently ignores them.</p>
    1426. 

  1426. </div>
    1427. 

  1427. </dd></dl>
    1428. 

  1428. 

  1429. </dd></dl>
    1430. 

  1430. 

  1431. <dl class="py class">
    1432. 

  1432. <dt class="sig sig-object py" id="super_gradients.training.losses.LabelSmoothingCrossEntropyLoss">
    1433. 

  1433. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">LabelSmoothingCrossEntropyLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'mean'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_dist</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">from_logits</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#LabelSmoothingCrossEntropyLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.LabelSmoothingCrossEntropyLoss" title="Permalink to this definition"></a></dt>
    1434. 

  1434. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss.CrossEntropyLoss</span></code></p>
    1435. 

  1435. <p>CrossEntropyLoss - with ability to recieve distrbution as targets, and optional label smoothing</p>
    1436. 

  1436. <dl class="py method">
    1437. 

  1437. <dt class="sig sig-object py" id="super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.forward">
    1438. 

  1438. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_dist</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/label_smoothing_cross_entropy_loss.html#LabelSmoothingCrossEntropyLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.forward" title="Permalink to this definition"></a></dt>
    1439. 

  1439. <dd><p>Defines the computation performed at every call.</p>
    1440. 

  1440. <p>Should be overridden by all subclasses.</p>
    1441. 

  1441. <div class="admonition note">
    1442. 

  1442. <p class="admonition-title">Note</p>
    1443. 

  1443. <p>Although the recipe for forward pass needs to be defined within
    1444. 

  1444. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1445. 

  1445. instead of this since the former takes care of running the
    1446. 

  1446. registered hooks while the latter silently ignores them.</p>
    1447. 

  1447. </div>
    1448. 

  1448. </dd></dl>
    1449. 

  1449. 

  1450. <dl class="py attribute">
    1451. 

  1451. <dt class="sig sig-object py" id="super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.ignore_index">
    1452. 

  1452. <span class="sig-name descname"><span class="pre">ignore_index</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.ignore_index" title="Permalink to this definition"></a></dt>
    1453. 

  1453. <dd></dd></dl>
    1454. 

  1454. 

  1455. <dl class="py attribute">
    1456. 

  1456. <dt class="sig sig-object py" id="super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.label_smoothing">
    1457. 

  1457. <span class="sig-name descname"><span class="pre">label_smoothing</span></span><em class="property"><span class="pre">:</span> <span class="pre">float</span></em><a class="headerlink" href="#super_gradients.training.losses.LabelSmoothingCrossEntropyLoss.label_smoothing" title="Permalink to this definition"></a></dt>
    1458. 

  1458. <dd></dd></dl>
    1459. 

  1459. 

  1460. </dd></dl>
    1461. 

  1461. 

  1462. <dl class="py class">
    1463. 

  1463. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetOHEMLoss">
    1464. 

  1464. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">ShelfNetOHEMLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.7</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mining_percent</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.0001</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_lb</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">255</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_ohem_loss.html#ShelfNetOHEMLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ShelfNetOHEMLoss" title="Permalink to this definition"></a></dt>
    1465. 

  1465. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.losses.html#super_gradients.training.losses.ohem_ce_loss.OhemCELoss" title="super_gradients.training.losses.ohem_ce_loss.OhemCELoss"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.losses.ohem_ce_loss.OhemCELoss</span></code></a></p>
    1466. 

  1466. <dl class="py method">
    1467. 

  1467. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetOHEMLoss.forward">
    1468. 

  1468. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions_list</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">list</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_ohem_loss.html#ShelfNetOHEMLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ShelfNetOHEMLoss.forward" title="Permalink to this definition"></a></dt>
    1469. 

  1469. <dd><p>Defines the computation performed at every call.</p>
    1470. 

  1470. <p>Should be overridden by all subclasses.</p>
    1471. 

  1471. <div class="admonition note">
    1472. 

  1472. <p class="admonition-title">Note</p>
    1473. 

  1473. <p>Although the recipe for forward pass needs to be defined within
    1474. 

  1474. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1475. 

  1475. instead of this since the former takes care of running the
    1476. 

  1476. registered hooks while the latter silently ignores them.</p>
    1477. 

  1477. </div>
    1478. 

  1478. </dd></dl>
    1479. 

  1479. 

  1480. <dl class="py attribute">
    1481. 

  1481. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetOHEMLoss.reduction">
    1482. 

  1482. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.ShelfNetOHEMLoss.reduction" title="Permalink to this definition"></a></dt>
    1483. 

  1483. <dd></dd></dl>
    1484. 

  1484. 

  1485. </dd></dl>
    1486. 

  1486. 

  1487. <dl class="py class">
    1488. 

  1488. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetSemanticEncodingLoss">
    1489. 

  1489. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">ShelfNetSemanticEncodingLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">se_weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">nclass</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">21</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">aux_weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.4</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ignore_index</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">1</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_semantic_encoding_loss.html#ShelfNetSemanticEncodingLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ShelfNetSemanticEncodingLoss" title="Permalink to this definition"></a></dt>
    1490. 

  1490. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss.CrossEntropyLoss</span></code></p>
    1491. 

  1491. <p>2D Cross Entropy Loss with Auxilary Loss</p>
    1492. 

  1492. <dl class="py method">
    1493. 

  1493. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetSemanticEncodingLoss.forward">
    1494. 

  1494. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">logits</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">labels</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/shelfnet_semantic_encoding_loss.html#ShelfNetSemanticEncodingLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.ShelfNetSemanticEncodingLoss.forward" title="Permalink to this definition"></a></dt>
    1495. 

  1495. <dd><p>Defines the computation performed at every call.</p>
    1496. 

  1496. <p>Should be overridden by all subclasses.</p>
    1497. 

  1497. <div class="admonition note">
    1498. 

  1498. <p class="admonition-title">Note</p>
    1499. 

  1499. <p>Although the recipe for forward pass needs to be defined within
    1500. 

  1500. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1501. 

  1501. instead of this since the former takes care of running the
    1502. 

  1502. registered hooks while the latter silently ignores them.</p>
    1503. 

  1503. </div>
    1504. 

  1504. </dd></dl>
    1505. 

  1505. 

  1506. <dl class="py attribute">
    1507. 

  1507. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetSemanticEncodingLoss.ignore_index">
    1508. 

  1508. <span class="sig-name descname"><span class="pre">ignore_index</span></span><em class="property"><span class="pre">:</span> <span class="pre">int</span></em><a class="headerlink" href="#super_gradients.training.losses.ShelfNetSemanticEncodingLoss.ignore_index" title="Permalink to this definition"></a></dt>
    1509. 

  1509. <dd></dd></dl>
    1510. 

  1510. 

  1511. <dl class="py attribute">
    1512. 

  1512. <dt class="sig sig-object py" id="super_gradients.training.losses.ShelfNetSemanticEncodingLoss.label_smoothing">
    1513. 

  1513. <span class="sig-name descname"><span class="pre">label_smoothing</span></span><em class="property"><span class="pre">:</span> <span class="pre">float</span></em><a class="headerlink" href="#super_gradients.training.losses.ShelfNetSemanticEncodingLoss.label_smoothing" title="Permalink to this definition"></a></dt>
    1514. 

  1514. <dd></dd></dl>
    1515. 

  1515. 

  1516. </dd></dl>
    1517. 

  1517. 

  1518. <dl class="py class">
    1519. 

  1519. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss">
    1520. 

  1520. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">YoloXDetectionLoss</span></span><span class="sig-paren">(</span><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></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">use_l1</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">center_sampling_radius</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">2.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">iou_type</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'iou'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss" title="Permalink to this definition"></a></dt>
    1521. 

  1521. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1522. 

  1522. <p>Calculate YOLOX loss:
    1523. 

  1523. L = L_objectivness + L_iou + L_classification + 1[use_l1]*L_l1</p>
    1524. 

  1524. <dl>
    1525. 

  1525. <dt>where:</dt><dd><ul class="simple">
    1526. 

  1526. <li><p>L_iou, L_classification and L_l1 are calculated only between cells and targets that suit them;</p></li>
    1527. 

  1527. <li><p>L_objectivness is calculated for all cells.</p></li>
    1528. 

  1528. </ul>
    1529. 

  1529. <dl class="simple">
    1530. 

  1530. <dt>L_classification:</dt><dd><p>for cells that have suitable ground truths in their grid locations add BCEs
    1531. 

  1531. to force a prediction of IoU with a GT in a multi-label way
    1532. 

  1532. Coef: 1.</p>
    1533. 

  1533. </dd>
    1534. 

  1534. <dt>L_iou:</dt><dd><p>for cells that have suitable ground truths in their grid locations
    1535. 

  1535. add (1 - IoU^2), IoU between a predicted box and each GT box, force maximum IoU
    1536. 

  1536. Coef: 5.</p>
    1537. 

  1537. </dd>
    1538. 

  1538. <dt>L_l1:</dt><dd><p>for cells that have suitable ground truths in their grid locations
    1539. 

  1539. l1 distance between the logits and GTs in “logits” format (the inverse of “logits to predictions” ops)
    1540. 

  1540. Coef: 1[use_l1]</p>
    1541. 

  1541. </dd>
    1542. 

  1542. <dt>L_objectness:</dt><dd><p>for each cell add BCE with a label of 1 if there is GT assigned to the cell
    1543. 

  1543. Coef: 1</p>
    1544. 

  1544. </dd>
    1545. 

  1545. </dl>
    1546. 

  1546. </dd>
    1547. 

  1547. </dl>
    1548. 

  1548. <dl class="py attribute">
    1549. 

  1549. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.strides">
    1550. 

  1550. <span class="sig-name descname"><span class="pre">strides</span></span><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.strides" title="Permalink to this definition"></a></dt>
    1551. 

  1551. <dd><p>list: List of Yolo levels output grid sizes (i.e [8, 16, 32]).</p>
    1552. 

  1552. </dd></dl>
    1553. 

  1553. 

  1554. <dl class="py attribute">
    1555. 

  1555. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.num_classes">
    1556. 

  1556. <span class="sig-name descname"><span class="pre">num_classes</span></span><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.num_classes" title="Permalink to this definition"></a></dt>
    1557. 

  1557. <dd><p>int: Number of classes.</p>
    1558. 

  1558. </dd></dl>
    1559. 

  1559. 

  1560. <dl class="py attribute">
    1561. 

  1561. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.use_l1">
    1562. 

  1562. <span class="sig-name descname"><span class="pre">use_l1</span></span><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.use_l1" title="Permalink to this definition"></a></dt>
    1563. 

  1563. <dd><p>bool: Controls the L_l1 Coef as discussed above (default=False).</p>
    1564. 

  1564. </dd></dl>
    1565. 

  1565. 

  1566. <dl class="py attribute">
    1567. 

  1567. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.center_sampling_radius">
    1568. 

  1568. <span class="sig-name descname"><span class="pre">center_sampling_radius</span></span><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.center_sampling_radius" title="Permalink to this definition"></a></dt>
    1569. 

  1569. <dd><p>float: Sampling radius used for center sampling when creating the fg mask (default=2.5).</p>
    1570. 

  1570. </dd></dl>
    1571. 

  1571. 

  1572. <dl class="py attribute">
    1573. 

  1573. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.iou_type">
    1574. 

  1574. <span class="sig-name descname"><span class="pre">iou_type</span></span><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.iou_type" title="Permalink to this definition"></a></dt>
    1575. 

  1575. <dd><p>str: Iou loss type, one of [“iou”,”giou”] (deafult=”iou”).</p>
    1576. 

  1576. </dd></dl>
    1577. 

  1577. 

  1578. <dl class="py method">
    1579. 

  1579. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.forward">
    1580. 

  1580. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">model_output</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">list</span><span class="p"><span class="pre">,</span> </span><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">List</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">targets</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.forward" title="Permalink to this definition"></a></dt>
    1581. 

  1581. <dd><dl class="field-list simple">
    1582. 

  1582. <dt class="field-odd">Parameters</dt>
    1583. 

  1583. <dd class="field-odd"><ul class="simple">
    1584. 

  1584. <li><p><strong>model_output</strong> – <p>Union[list, Tuple[torch.Tensor, List]]:
    1585. 

  1585. When list-</p>
    1586. 

  1586. <blockquote>
    1587. 

  1587. <div><p>output from all Yolo levels, each of shape [Batch x 1 x GridSizeY x GridSizeX x (4 + 1 + Num_classes)]</p>
    1588. 

  1588. </div></blockquote>
    1589. 

  1589. <p>And when tuple- the second item is the described list (first item is discarded)</p>
    1590. 

  1590. </p></li>
    1591. 

  1591. <li><p><strong>targets</strong> – torch.Tensor: Num_targets x (4 + 2)], values on dim 1 are: image id in a batch, class, box x y w h</p></li>
    1592. 

  1592. </ul>
    1593. 

  1593. </dd>
    1594. 

  1594. <dt class="field-even">Returns</dt>
    1595. 

  1595. <dd class="field-even"><p>loss, all losses separately in a detached tensor</p>
    1596. 

  1596. </dd>
    1597. 

  1597. </dl>
    1598. 

  1598. </dd></dl>
    1599. 

  1599. 

  1600. <dl class="py method">
    1601. 

  1601. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.prepare_predictions">
    1602. 

  1602. <span class="sig-name descname"><span class="pre">prepare_predictions</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em><span class="sig-paren">)</span> &#x2192; <span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.prepare_predictions"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.prepare_predictions" title="Permalink to this definition"></a></dt>
    1603. 

  1603. <dd><p>Convert raw outputs of the network into a format that merges outputs from all levels
    1604. 

  1604. :param predictions:     output from all Yolo levels, each of shape</p>
    1605. 

  1605. <blockquote>
    1606. 

  1606. <div><p>[Batch x 1 x GridSizeY x GridSizeX x (4 + 1 + Num_classes)]</p>
    1607. 

  1607. </div></blockquote>
    1608. 

  1608. <dl class="field-list simple">
    1609. 

  1609. <dt class="field-odd">Returns</dt>
    1610. 

  1610. <dd class="field-odd"><p><p>5 tensors representing predictions:
    1611. 

  1611. * x_shifts: shape [1 x * num_cells x 1],</p>
    1612. 

  1612. <blockquote>
    1613. 

  1613. <div><p>where num_cells = grid1X * grid1Y + grid2X * grid2Y + grid3X * grid3Y,
    1614. 

  1614. x coordinate on the grid cell the prediction is coming from</p>
    1615. 

  1615. </div></blockquote>
    1616. 

  1616. <ul class="simple">
    1617. 

  1617. <li><p>y_shifts: shape [1 x  num_cells x 1],
    1618. 

  1618. y coordinate on the grid cell the prediction is coming from</p></li>
    1619. 

  1619. <li><p>expanded_strides: shape [1 x num_cells x 1],
    1620. 

  1620. stride of the output grid the prediction is coming from</p></li>
    1621. 

  1621. <li><p>transformed_outputs: shape [batch_size x num_cells x (num_classes + 5)],
    1622. 

  1622. predictions with boxes in real coordinates and logprobabilities</p></li>
    1623. 

  1623. <li><p>raw_outputs: shape [batch_size x num_cells x (num_classes + 5)],
    1624. 

  1624. raw predictions with boxes and confidences as logits</p></li>
    1625. 

  1625. </ul>
    1626. 

  1626. </p>
    1627. 

  1627. </dd>
    1628. 

  1628. </dl>
    1629. 

  1629. </dd></dl>
    1630. 

  1630. 

  1631. <dl class="py method">
    1632. 

  1632. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.get_l1_target">
    1633. 

  1633. <span class="sig-name descname"><span class="pre">get_l1_target</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">l1_target</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gt</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stride</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">x_shifts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_shifts</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-08</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.get_l1_target"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.get_l1_target" title="Permalink to this definition"></a></dt>
    1634. 

  1634. <dd><dl class="field-list simple">
    1635. 

  1635. <dt class="field-odd">Parameters</dt>
    1636. 

  1636. <dd class="field-odd"><ul class="simple">
    1637. 

  1637. <li><p><strong>l1_target</strong> – tensor of zeros of shape [Num_cell_gt_pairs x 4]</p></li>
    1638. 

  1638. <li><p><strong>gt</strong> – targets in coordinates [Num_cell_gt_pairs x (4 + 1 + num_classes)]</p></li>
    1639. 

  1639. </ul>
    1640. 

  1640. </dd>
    1641. 

  1641. <dt class="field-even">Returns</dt>
    1642. 

  1642. <dd class="field-even"><p>targets in the format corresponding to logits</p>
    1643. 

  1643. </dd>
    1644. 

  1644. </dl>
    1645. 

  1645. </dd></dl>
    1646. 

  1646. 

  1647. <dl class="py method">
    1648. 

  1648. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.get_assignments">
    1649. 

  1649. <span class="sig-name descname"><span class="pre">get_assignments</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">image_idx</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_gt</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">total_num_anchors</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gt_bboxes_per_image</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gt_classes</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">bboxes_preds_per_image</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">expanded_strides</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">x_shifts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_shifts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">cls_preds</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">obj_preds</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">mode</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'gpu'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">ious_loss_cost_coeff</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">outside_boxes_and_center_cost_coeff</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">100000.0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.get_assignments"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.get_assignments" title="Permalink to this definition"></a></dt>
    1650. 

  1650. <dd><dl class="simple">
    1651. 

  1651. <dt>Match cells to ground truth:</dt><dd><ul class="simple">
    1652. 

  1652. <li><p>at most 1 GT per cell</p></li>
    1653. 

  1653. <li><p>dynamic number of cells per GT</p></li>
    1654. 

  1654. </ul>
    1655. 

  1655. </dd>
    1656. 

  1656. </dl>
    1657. 

  1657. <dl class="field-list simple">
    1658. 

  1658. <dt class="field-odd">Parameters</dt>
    1659. 

  1659. <dd class="field-odd"><ul class="simple">
    1660. 

  1660. <li><p><strong>outside_boxes_and_center_cost_coeff</strong> – float: Cost coefficiant of cells the radius and bbox of gts in dynamic
    1661. 

  1661. matching (default=100000).</p></li>
    1662. 

  1662. <li><p><strong>ious_loss_cost_coeff</strong> – float: Cost coefficiant for iou loss in dynamic matching (default=3).</p></li>
    1663. 

  1663. <li><p><strong>image_idx</strong> – int: Image index in batch.</p></li>
    1664. 

  1664. <li><p><strong>num_gt</strong> – int: Number of ground trunth targets in the image.</p></li>
    1665. 

  1665. <li><p><strong>total_num_anchors</strong> – int: Total number of possible bboxes = sum of all grid cells.</p></li>
    1666. 

  1666. <li><p><strong>gt_bboxes_per_image</strong> – torch.Tensor: Tensor of gt bboxes for  the image, shape: (num_gt, 4).</p></li>
    1667. 

  1667. <li><p><strong>gt_classes</strong> – torch.Tesnor: Tensor of the classes in the image, shape: (num_preds,4).</p></li>
    1668. 

  1668. <li><p><strong>bboxes_preds_per_image</strong> – Tensor of the classes in the image, shape: (num_preds).</p></li>
    1669. 

  1669. <li><p><strong>expanded_strides</strong> – torch.Tensor: Stride of the output grid the prediction is coming from,
    1670. 

  1670. shape (1 x num_cells x 1).</p></li>
    1671. 

  1671. <li><p><strong>x_shifts</strong> – torch.Tensor: X’s in cell coordinates, shape (1,num_cells,1).</p></li>
    1672. 

  1672. <li><p><strong>y_shifts</strong> – torch.Tensor: Y’s in cell coordinates, shape (1,num_cells,1).</p></li>
    1673. 

  1673. <li><p><strong>cls_preds</strong> – torch.Tensor: Class predictions in all cells, shape (batch_size, num_cells).</p></li>
    1674. 

  1674. <li><p><strong>obj_preds</strong> – torch.Tensor: Objectness predictions in all cells, shape (batch_size, num_cells).</p></li>
    1675. 

  1675. <li><p><strong>mode</strong> – str: One of [“gpu”,”cpu”], Controls the device the assignment operation should be taken place on (deafult=”gpu”)</p></li>
    1676. 

  1676. </ul>
    1677. 

  1677. </dd>
    1678. 

  1678. </dl>
    1679. 

  1679. </dd></dl>
    1680. 

  1680. 

  1681. <dl class="py method">
    1682. 

  1682. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.get_in_boxes_info">
    1683. 

  1683. <span class="sig-name descname"><span class="pre">get_in_boxes_info</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">gt_bboxes_per_image</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">expanded_strides</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">x_shifts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_shifts</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">total_num_anchors</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_gt</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.get_in_boxes_info"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.get_in_boxes_info" title="Permalink to this definition"></a></dt>
    1684. 

  1684. <dd><dl>
    1685. 

  1685. <dt>Create a mask for all cells, mask in only foreground: cells that have a center located:</dt><dd><ul class="simple">
    1686. 

  1686. <li><p>withing a GT box;</p></li>
    1687. 

  1687. </ul>
    1688. 

  1688. <p>OR
    1689. 

  1689. * within a fixed radius around a GT box (center sampling);</p>
    1690. 

  1690. </dd>
    1691. 

  1691. </dl>
    1692. 

  1692. <dl class="field-list simple">
    1693. 

  1693. <dt class="field-odd">Parameters</dt>
    1694. 

  1694. <dd class="field-odd"><ul class="simple">
    1695. 

  1695. <li><p><strong>num_gt</strong> – int: Number of ground trunth targets in the image.</p></li>
    1696. 

  1696. <li><p><strong>total_num_anchors</strong> – int: Sum of all grid cells.</p></li>
    1697. 

  1697. <li><p><strong>gt_bboxes_per_image</strong> – torch.Tensor: Tensor of gt bboxes for  the image, shape: (num_gt, 4).</p></li>
    1698. 

  1698. <li><p><strong>expanded_strides</strong> – torch.Tensor: Stride of the output grid the prediction is coming from,
    1699. 

  1699. shape (1 x num_cells x 1).</p></li>
    1700. 

  1700. <li><p><strong>x_shifts</strong> – torch.Tensor: X’s in cell coordinates, shape (1,num_cells,1).</p></li>
    1701. 

  1701. <li><p><strong>y_shifts</strong> – torch.Tensor: Y’s in cell coordinates, shape (1,num_cells,1).</p></li>
    1702. 

  1702. </ul>
    1703. 

  1703. </dd>
    1704. 

  1704. </dl>
    1705. 

  1705. <dl class="simple">
    1706. 

  1706. <dt>:return is_in_boxes_anchor, is_in_boxes_and_center</dt><dd><dl class="simple">
    1707. 

  1707. <dt>where:</dt><dd><ul class="simple">
    1708. 

  1708. <li><dl class="simple">
    1709. 

  1709. <dt>is_in_boxes_anchor masks the cells that their cell center is  inside a gt bbox and within</dt><dd><p>self.center_sampling_radius cells away, without reduction (i.e shape=(num_gts, num_fgs))</p>
    1710. 

  1710. </dd>
    1711. 

  1711. </dl>
    1712. 

  1712. </li>
    1713. 

  1713. <li><dl class="simple">
    1714. 

  1714. <dt>is_in_boxes_and_center masks the cells that their center is either inside a gt bbox or within</dt><dd><p>self.center_sampling_radius cells away, shape (num_fgs)</p>
    1715. 

  1715. </dd>
    1716. 

  1716. </dl>
    1717. 

  1717. </li>
    1718. 

  1718. </ul>
    1719. 

  1719. </dd>
    1720. 

  1720. </dl>
    1721. 

  1721. </dd>
    1722. 

  1722. </dl>
    1723. 

  1723. </dd></dl>
    1724. 

  1724. 

  1725. <dl class="py method">
    1726. 

  1726. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.dynamic_k_matching">
    1727. 

  1727. <span class="sig-name descname"><span class="pre">dynamic_k_matching</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">cost</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">pair_wise_ious</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gt_classes</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_gt</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fg_mask</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/yolox_loss.html#YoloXDetectionLoss.dynamic_k_matching"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.dynamic_k_matching" title="Permalink to this definition"></a></dt>
    1728. 

  1728. <dd><dl class="field-list simple">
    1729. 

  1729. <dt class="field-odd">Parameters</dt>
    1730. 

  1730. <dd class="field-odd"><ul class="simple">
    1731. 

  1731. <li><p><strong>cost</strong> – pairwise cost, [num_FGs x num_GTs]</p></li>
    1732. 

  1732. <li><p><strong>pair_wise_ious</strong> – pairwise IoUs, [num_FGs x num_GTs]</p></li>
    1733. 

  1733. <li><p><strong>gt_classes</strong> – class of each GT</p></li>
    1734. 

  1734. <li><p><strong>num_gt</strong> – number of GTs</p></li>
    1735. 

  1735. </ul>
    1736. 

  1736. </dd>
    1737. 

  1737. </dl>
    1738. 

  1738. <dl class="simple">
    1739. 

  1739. <dt>:return num_fg, (number of foregrounds)</dt><dd><p>gt_matched_classes, (the classes that have been matched with fgs)
    1740. 

  1740. pred_ious_this_matching
    1741. 

  1741. matched_gt_inds</p>
    1742. 

  1742. </dd>
    1743. 

  1743. </dl>
    1744. 

  1744. </dd></dl>
    1745. 

  1745. 

  1746. <dl class="py attribute">
    1747. 

  1747. <dt class="sig sig-object py" id="super_gradients.training.losses.YoloXDetectionLoss.reduction">
    1748. 

  1748. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.YoloXDetectionLoss.reduction" title="Permalink to this definition"></a></dt>
    1749. 

  1749. <dd></dd></dl>
    1750. 

  1750. 

  1751. </dd></dl>
    1752. 

  1752. 

  1753. <dl class="py class">
    1754. 

  1754. <dt class="sig sig-object py" id="super_gradients.training.losses.RSquaredLoss">
    1755. 

  1755. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">RSquaredLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">size_average</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduce</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'mean'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/r_squared_loss.html#RSquaredLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.RSquaredLoss" title="Permalink to this definition"></a></dt>
    1756. 

  1756. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1757. 

  1757. <dl class="py method">
    1758. 

  1758. <dt class="sig sig-object py" id="super_gradients.training.losses.RSquaredLoss.forward">
    1759. 

  1759. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/r_squared_loss.html#RSquaredLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.RSquaredLoss.forward" title="Permalink to this definition"></a></dt>
    1760. 

  1760. <dd><p>Computes the R-squared for the output and target values
    1761. 

  1761. :param output: Tensor / Numpy / List</p>
    1762. 

  1762. <blockquote>
    1763. 

  1763. <div><p>The prediction</p>
    1764. 

  1764. </div></blockquote>
    1765. 

  1765. <dl class="field-list simple">
    1766. 

  1766. <dt class="field-odd">Parameters</dt>
    1767. 

  1767. <dd class="field-odd"><p><strong>target</strong> – Tensor / Numpy / List
    1768. 

  1768. The corresponding lables</p>
    1769. 

  1769. </dd>
    1770. 

  1770. </dl>
    1771. 

  1771. </dd></dl>
    1772. 

  1772. 

  1773. <dl class="py attribute">
    1774. 

  1774. <dt class="sig sig-object py" id="super_gradients.training.losses.RSquaredLoss.reduction">
    1775. 

  1775. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.RSquaredLoss.reduction" title="Permalink to this definition"></a></dt>
    1776. 

  1776. <dd></dd></dl>
    1777. 

  1777. 

  1778. </dd></dl>
    1779. 

  1779. 

  1780. <dl class="py class">
    1781. 

  1781. <dt class="sig sig-object py" id="super_gradients.training.losses.SSDLoss">
    1782. 

  1782. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">SSDLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dboxes</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.utils.html#super_gradients.training.utils.ssd_utils.DefaultBoxes" title="super_gradients.training.utils.ssd_utils.DefaultBoxes"><span class="pre">super_gradients.training.utils.ssd_utils.DefaultBoxes</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">alpha</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">1.0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">iou_thresh</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">neg_pos_ratio</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">3.0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.SSDLoss" title="Permalink to this definition"></a></dt>
    1783. 

  1783. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1784. 

  1784. <blockquote>
    1785. 

  1785. <div><p>Implements the loss as the sum of the followings:
    1786. 

  1786. 1. Confidence Loss: All labels, with hard negative mining
    1787. 

  1787. 2. Localization Loss: Only on positive labels</p>
    1788. 

  1788. </div></blockquote>
    1789. 

  1789. <dl class="simple">
    1790. 

  1790. <dt>L = (2 - alpha) * L_l1 + alpha * L_cls, where</dt><dd><ul class="simple">
    1791. 

  1791. <li><p>L_cls is HardMiningCrossEntropyLoss</p></li>
    1792. 

  1792. <li><p>L_l1 = [SmoothL1Loss for all positives]</p></li>
    1793. 

  1793. </ul>
    1794. 

  1794. </dd>
    1795. 

  1795. </dl>
    1796. 

  1796. <dl class="py method">
    1797. 

  1797. <dt class="sig sig-object py" id="super_gradients.training.losses.SSDLoss.match_dboxes">
    1798. 

  1798. <span class="sig-name descname"><span class="pre">match_dboxes</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss.match_dboxes"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.SSDLoss.match_dboxes" title="Permalink to this definition"></a></dt>
    1799. 

  1799. <dd><p>creates tensors with target boxes and labels for each dboxes, so with the same len as dboxes.</p>
    1800. 

  1800. <ul class="simple">
    1801. 

  1801. <li><p>Each GT is assigned with a grid cell with the highest IoU, this creates a pair for each GT and some cells;</p></li>
    1802. 

  1802. <li><p>The rest of grid cells are assigned to a GT with the highest IoU, assuming it’s &gt; self.iou_thresh;
    1803. 

  1803. If this condition is not met the grid cell is marked as background</p></li>
    1804. 

  1804. </ul>
    1805. 

  1805. <p>GT-wise: one to many
    1806. 

  1806. Grid-cell-wise: one to one</p>
    1807. 

  1807. <dl class="field-list simple">
    1808. 

  1808. <dt class="field-odd">Parameters</dt>
    1809. 

  1809. <dd class="field-odd"><p><strong>targets</strong> – a tensor containing the boxes for a single image;
    1810. 

  1810. shape [num_boxes, 6] (image_id, label, x, y, w, h)</p>
    1811. 

  1811. </dd>
    1812. 

  1812. <dt class="field-even">Returns</dt>
    1813. 

  1813. <dd class="field-even"><p>two tensors
    1814. 

  1814. boxes - shape of dboxes [4, num_dboxes] (x,y,w,h)
    1815. 

  1815. labels - sahpe [num_dboxes]</p>
    1816. 

  1816. </dd>
    1817. 

  1817. </dl>
    1818. 

  1818. </dd></dl>
    1819. 

  1819. 

  1820. <dl class="py method">
    1821. 

  1821. <dt class="sig sig-object py" id="super_gradients.training.losses.SSDLoss.forward">
    1822. 

  1822. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">predictions</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Tuple</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/ssd_loss.html#SSDLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.SSDLoss.forward" title="Permalink to this definition"></a></dt>
    1823. 

  1823. <dd><dl class="simple">
    1824. 

  1824. <dt>Compute the loss</dt><dd><p>:param predictions - predictions tensor coming from the network,
    1825. 

  1825. tuple with shapes ([Batch Size, 4, num_dboxes], [Batch Size, num_classes + 1, num_dboxes])
    1826. 

  1826. were predictions have logprobs for background and other classes
    1827. 

  1827. :param targets - targets for the batch. [num targets, 6] (index in batch, label, x,y,w,h)</p>
    1828. 

  1828. </dd>
    1829. 

  1829. </dl>
    1830. 

  1830. </dd></dl>
    1831. 

  1831. 

  1832. <dl class="py attribute">
    1833. 

  1833. <dt class="sig sig-object py" id="super_gradients.training.losses.SSDLoss.reduction">
    1834. 

  1834. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.SSDLoss.reduction" title="Permalink to this definition"></a></dt>
    1835. 

  1835. <dd></dd></dl>
    1836. 

  1836. 

  1837. </dd></dl>
    1838. 

  1838. 

  1839. <dl class="py class">
    1840. 

  1840. <dt class="sig sig-object py" id="super_gradients.training.losses.BCEDiceLoss">
    1841. 

  1841. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">BCEDiceLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">loss_weights</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">[0.5,</span> <span class="pre">0.5]</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">logits</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/bce_dice_loss.html#BCEDiceLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.BCEDiceLoss" title="Permalink to this definition"></a></dt>
    1842. 

  1842. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    1843. 

  1843. <p>Binary Cross Entropy + Dice Loss</p>
    1844. 

  1844. <p>Weighted average of BCE and Dice loss</p>
    1845. 

  1845. <dl class="py attribute">
    1846. 

  1846. <dt class="sig sig-object py" id="super_gradients.training.losses.BCEDiceLoss.loss_weights">
    1847. 

  1847. <span class="sig-name descname"><span class="pre">loss_weights</span></span><a class="headerlink" href="#super_gradients.training.losses.BCEDiceLoss.loss_weights" title="Permalink to this definition"></a></dt>
    1848. 

  1848. <dd><p>list of size 2 s.t loss_weights[0], loss_weights[1] are the weights for BCE, Dice</p>
    1849. 

  1849. </dd></dl>
    1850. 

  1850. 

  1851. <dl class="py attribute">
    1852. 

  1852. <dt class="sig sig-object py">
    1853. 

  1853. <span class="sig-name descname"><span class="pre">respectively.</span></span></dt>
    1854. 

  1854. <dd></dd></dl>
    1855. 

  1855. 

  1856. <dl class="py method">
    1857. 

  1857. <dt class="sig sig-object py" id="super_gradients.training.losses.BCEDiceLoss.forward">
    1858. 

  1858. <span class="sig-name descname"><span class="pre">forward</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">input</span></span><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">target</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">torch.Tensor</span><a class="reference internal" href="_modules/super_gradients/training/losses/bce_dice_loss.html#BCEDiceLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.BCEDiceLoss.forward" title="Permalink to this definition"></a></dt>
    1859. 

  1859. <dd><p>&#64;param input: Network’s raw output shaped (N,1,H,W)
    1860. 

  1860. &#64;param target: Ground truth shaped (N,H,W)</p>
    1861. 

  1861. </dd></dl>
    1862. 

  1862. 

  1863. <dl class="py attribute">
    1864. 

  1864. <dt class="sig sig-object py" id="super_gradients.training.losses.BCEDiceLoss.training">
    1865. 

  1865. <span class="sig-name descname"><span class="pre">training</span></span><em class="property"><span class="pre">:</span> <span class="pre">bool</span></em><a class="headerlink" href="#super_gradients.training.losses.BCEDiceLoss.training" title="Permalink to this definition"></a></dt>
    1866. 

  1866. <dd></dd></dl>
    1867. 

  1867. 

  1868. </dd></dl>
    1869. 

  1869. 

  1870. <dl class="py class">
    1871. 

  1871. <dt class="sig sig-object py" id="super_gradients.training.losses.KDLogitsLoss">
    1872. 

  1872. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">KDLogitsLoss</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task_loss_fn</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.loss._Loss</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">distillation_loss_fn</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.nn.modules.loss._Loss</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">KDklDivLoss()</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">distillation_loss_coeff</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/losses/kd_losses.html#KDLogitsLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.KDLogitsLoss" title="Permalink to this definition"></a></dt>
    1873. 

  1873. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1874. 

  1874. <p>Knowledge distillation loss, wraps the task loss and distillation loss</p>
    1875. 

  1875. <dl class="py method">
    1876. 

  1876. <dt class="sig sig-object py" id="super_gradients.training.losses.KDLogitsLoss.forward">
    1877. 

  1877. <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">kd_module_output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/kd_losses.html#KDLogitsLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.KDLogitsLoss.forward" title="Permalink to this definition"></a></dt>
    1878. 

  1878. <dd><p>Defines the computation performed at every call.</p>
    1879. 

  1879. <p>Should be overridden by all subclasses.</p>
    1880. 

  1880. <div class="admonition note">
    1881. 

  1881. <p class="admonition-title">Note</p>
    1882. 

  1882. <p>Although the recipe for forward pass needs to be defined within
    1883. 

  1883. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    1884. 

  1884. instead of this since the former takes care of running the
    1885. 

  1885. registered hooks while the latter silently ignores them.</p>
    1886. 

  1886. </div>
    1887. 

  1887. </dd></dl>
    1888. 

  1888. 

  1889. <dl class="py attribute">
    1890. 

  1890. <dt class="sig sig-object py" id="super_gradients.training.losses.KDLogitsLoss.reduction">
    1891. 

  1891. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.KDLogitsLoss.reduction" title="Permalink to this definition"></a></dt>
    1892. 

  1892. <dd></dd></dl>
    1893. 

  1893. 

  1894. </dd></dl>
    1895. 

  1895. 

  1896. <dl class="py class">
    1897. 

  1897. <dt class="sig sig-object py" id="super_gradients.training.losses.DiceCEEdgeLoss">
    1898. 

  1898. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.losses.</span></span><span class="sig-name descname"><span class="pre">DiceCEEdgeLoss</span></span><span class="sig-paren">(</span><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">num_aux_heads</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">2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">num_detail_heads</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">weights</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">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">(1,</span> <span class="pre">1,</span> <span class="pre">1,</span> <span class="pre">1)</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dice_ce_weights</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">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">(1,</span> <span class="pre">1)</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">int</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">edge_kernel</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">ce_edge_weights</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">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">0.5)</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/dice_ce_edge_loss.html#DiceCEEdgeLoss"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.DiceCEEdgeLoss" title="Permalink to this definition"></a></dt>
    1899. 

  1899. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.loss._Loss</span></code></p>
    1900. 

  1900. <dl class="py method">
    1901. 

  1901. <dt class="sig sig-object py" id="super_gradients.training.losses.DiceCEEdgeLoss.forward">
    1902. 

  1902. <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">preds</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Tuple</span><span class="p"><span class="pre">[</span></span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch.Tensor</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/losses/dice_ce_edge_loss.html#DiceCEEdgeLoss.forward"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.losses.DiceCEEdgeLoss.forward" title="Permalink to this definition"></a></dt>
    1903. 

  1903. <dd><dl class="field-list simple">
    1904. 

  1904. <dt class="field-odd">Parameters</dt>
    1905. 

  1905. <dd class="field-odd"><p><strong>preds</strong> – Model output predictions, must be in the followed format:
    1906. 

  1906. [Main-feats, Aux-feats[0], …, Aux-feats[num_auxs-1], Detail-feats[0], …, Detail-feats[num_details-1]</p>
    1907. 

  1907. </dd>
    1908. 

  1908. </dl>
    1909. 

  1909. </dd></dl>
    1910. 

  1910. 

  1911. <dl class="py attribute">
    1912. 

  1912. <dt class="sig sig-object py" id="super_gradients.training.losses.DiceCEEdgeLoss.reduction">
    1913. 

  1913. <span class="sig-name descname"><span class="pre">reduction</span></span><em class="property"><span class="pre">:</span> <span class="pre">str</span></em><a class="headerlink" href="#super_gradients.training.losses.DiceCEEdgeLoss.reduction" title="Permalink to this definition"></a></dt>
    1914. 

  1914. <dd></dd></dl>
    1915. 

  1915. 

  1916. </dd></dl>
    1917. 

  1917. 

  1918. </section>
    1919. 

  1919. <section id="module-super_gradients.training.metrics">
    1920. 

  1920. <span id="super-gradients-training-metrics-module"></span><h2>super_gradients.training.metrics module<a class="headerlink" href="#module-super_gradients.training.metrics" title="Permalink to this headline"></a></h2>
    1921. 

  1921. <dl class="py function">
    1922. 

  1922. <dt class="sig sig-object py" id="super_gradients.training.metrics.accuracy">
    1923. 

  1923. <span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">accuracy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">output</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">target</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">topk</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/metrics/classification_metrics.html#accuracy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.accuracy" title="Permalink to this definition"></a></dt>
    1924. 

  1924. <dd><p>Computes the precision&#64;k for the specified values of k
    1925. 

  1925. :param output: Tensor / Numpy / List</p>
    1926. 

  1926. <blockquote>
    1927. 

  1927. <div><p>The prediction</p>
    1928. 

  1928. </div></blockquote>
    1929. 

  1929. <dl class="field-list simple">
    1930. 

  1930. <dt class="field-odd">Parameters</dt>
    1931. 

  1931. <dd class="field-odd"><ul class="simple">
    1932. 

  1932. <li><p><strong>target</strong> – Tensor / Numpy / List
    1933. 

  1933. The corresponding lables</p></li>
    1934. 

  1934. <li><p><strong>topk</strong> – tuple
    1935. 

  1935. The type of accuracy to calculate, e.g. topk=(1,5) returns accuracy for top-1 and top-5</p></li>
    1936. 

  1936. </ul>
    1937. 

  1937. </dd>
    1938. 

  1938. </dl>
    1939. 

  1939. </dd></dl>
    1940. 

  1940. 

  1941. <dl class="py class">
    1942. 

  1942. <dt class="sig sig-object py" id="super_gradients.training.metrics.Accuracy">
    1943. 

  1943. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">Accuracy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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/metrics/classification_metrics.html#Accuracy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Accuracy" title="Permalink to this definition"></a></dt>
    1944. 

  1944. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.classification.accuracy.Accuracy</span></code></p>
    1945. 

  1945. <dl class="py method">
    1946. 

  1946. <dt class="sig sig-object py" id="super_gradients.training.metrics.Accuracy.update">
    1947. 

  1947. <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">preds</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">target</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/metrics/classification_metrics.html#Accuracy.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Accuracy.update" title="Permalink to this definition"></a></dt>
    1948. 

  1948. <dd><p>Update state with predictions and targets. See
    1949. 

  1949. <span class="xref std std-ref">pages/classification:input types</span> for more information on input
    1950. 

  1950. types.</p>
    1951. 

  1951. <dl class="field-list simple">
    1952. 

  1952. <dt class="field-odd">Parameters</dt>
    1953. 

  1953. <dd class="field-odd"><ul class="simple">
    1954. 

  1954. <li><p><strong>preds</strong> – Predictions from model (logits, probabilities, or labels)</p></li>
    1955. 

  1955. <li><p><strong>target</strong> – Ground truth labels</p></li>
    1956. 

  1956. </ul>
    1957. 

  1957. </dd>
    1958. 

  1958. </dl>
    1959. 

  1959. </dd></dl>
    1960. 

  1960. 

  1961. <dl class="py attribute">
    1962. 

  1962. <dt class="sig sig-object py" id="super_gradients.training.metrics.Accuracy.correct">
    1963. 

  1963. <span class="sig-name descname"><span class="pre">correct</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.Accuracy.correct" title="Permalink to this definition"></a></dt>
    1964. 

  1964. <dd></dd></dl>
    1965. 

  1965. 

  1966. <dl class="py attribute">
    1967. 

  1967. <dt class="sig sig-object py" id="super_gradients.training.metrics.Accuracy.total">
    1968. 

  1968. <span class="sig-name descname"><span class="pre">total</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.Accuracy.total" title="Permalink to this definition"></a></dt>
    1969. 

  1969. <dd></dd></dl>
    1970. 

  1970. 

  1971. </dd></dl>
    1972. 

  1972. 

  1973. <dl class="py class">
    1974. 

  1974. <dt class="sig sig-object py" id="super_gradients.training.metrics.Top5">
    1975. 

  1975. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">Top5</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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/metrics/classification_metrics.html#Top5"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Top5" title="Permalink to this definition"></a></dt>
    1976. 

  1976. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.metric.Metric</span></code></p>
    1977. 

  1977. <dl class="py method">
    1978. 

  1978. <dt class="sig sig-object py" id="super_gradients.training.metrics.Top5.update">
    1979. 

  1979. <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">preds</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">target</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/metrics/classification_metrics.html#Top5.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Top5.update" title="Permalink to this definition"></a></dt>
    1980. 

  1980. <dd><p>Override this method to update the state variables of your metric class.</p>
    1981. 

  1981. </dd></dl>
    1982. 

  1982. 

  1983. <dl class="py method">
    1984. 

  1984. <dt class="sig sig-object py" id="super_gradients.training.metrics.Top5.compute">
    1985. 

  1985. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/classification_metrics.html#Top5.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Top5.compute" title="Permalink to this definition"></a></dt>
    1986. 

  1986. <dd><p>Override this method to compute the final metric value from state variables synchronized across the
    1987. 

  1987. distributed backend.</p>
    1988. 

  1988. </dd></dl>
    1989. 

  1989. 

  1990. </dd></dl>
    1991. 

  1991. 

  1992. <dl class="py class">
    1993. 

  1993. <dt class="sig sig-object py" id="super_gradients.training.metrics.ToyTestClassificationMetric">
    1994. 

  1994. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">ToyTestClassificationMetric</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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/metrics/classification_metrics.html#ToyTestClassificationMetric"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.ToyTestClassificationMetric" title="Permalink to this definition"></a></dt>
    1995. 

  1995. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.metric.Metric</span></code></p>
    1996. 

  1996. <p>Dummy classification Mettric object returning 0 always (for testing).</p>
    1997. 

  1997. <dl class="py method">
    1998. 

  1998. <dt class="sig sig-object py" id="super_gradients.training.metrics.ToyTestClassificationMetric.update">
    1999. 

  1999. <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">preds</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">target</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">None</span><a class="reference internal" href="_modules/super_gradients/training/metrics/classification_metrics.html#ToyTestClassificationMetric.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.ToyTestClassificationMetric.update" title="Permalink to this definition"></a></dt>
    2000. 

  2000. <dd><p>Override this method to update the state variables of your metric class.</p>
    2001. 

  2001. </dd></dl>
    2002. 

  2002. 

  2003. <dl class="py method">
    2004. 

  2004. <dt class="sig sig-object py" id="super_gradients.training.metrics.ToyTestClassificationMetric.compute">
    2005. 

  2005. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/classification_metrics.html#ToyTestClassificationMetric.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.ToyTestClassificationMetric.compute" title="Permalink to this definition"></a></dt>
    2006. 

  2006. <dd><p>Override this method to compute the final metric value from state variables synchronized across the
    2007. 

  2007. distributed backend.</p>
    2008. 

  2008. </dd></dl>
    2009. 

  2009. 

  2010. </dd></dl>
    2011. 

  2011. 

  2012. <dl class="py class">
    2013. 

  2013. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics">
    2014. 

  2014. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">DetectionMetrics</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">num_cls:</span> <span class="pre">int</span></em>, <em class="sig-param"><span class="pre">post_prediction_callback:</span> <span class="pre">Optional[super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">normalize_targets:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></em>, <em class="sig-param"><span class="pre">iou_thres:</span> <span class="pre">super_gradients.training.utils.detection_utils.IouThreshold</span> <span class="pre">=</span> <span class="pre">&lt;IouThreshold.MAP_05_TO_095:</span> <span class="pre">(0.5</span></em>, <em class="sig-param"><span class="pre">0.95)&gt;</span></em>, <em class="sig-param"><span class="pre">recall_thres:</span> <span class="pre">Optional[torch.Tensor]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">score_thres:</span> <span class="pre">float</span> <span class="pre">=</span> <span class="pre">0.1</span></em>, <em class="sig-param"><span class="pre">top_k_predictions:</span> <span class="pre">int</span> <span class="pre">=</span> <span class="pre">100</span></em>, <em class="sig-param"><span class="pre">dist_sync_on_step:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></em>, <em class="sig-param"><span class="pre">accumulate_on_cpu:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/detection_metrics.html#DetectionMetrics"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics" title="Permalink to this definition"></a></dt>
    2015. 

  2015. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.metric.Metric</span></code></p>
    2016. 

  2016. <p>Metric class for computing F1, Precision, Recall and Mean Average Precision.</p>
    2017. 

  2017. <dl class="py attribute">
    2018. 

  2018. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.num_cls">
    2019. 

  2019. <span class="sig-name descname"><span class="pre">num_cls</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.num_cls" title="Permalink to this definition"></a></dt>
    2020. 

  2020. <dd><p>Number of classes.</p>
    2021. 

  2021. </dd></dl>
    2022. 

  2022. 

  2023. <dl class="py attribute">
    2024. 

  2024. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.post_prediction_callback">
    2025. 

  2025. <span class="sig-name descname"><span class="pre">post_prediction_callback</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.post_prediction_callback" title="Permalink to this definition"></a></dt>
    2026. 

  2026. <dd><p>DetectionPostPredictionCallback to be applied on net’s output prior
    2027. 

  2027. to the metric computation (NMS).</p>
    2028. 

  2028. </dd></dl>
    2029. 

  2029. 

  2030. <dl class="py attribute">
    2031. 

  2031. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.normalize_targets">
    2032. 

  2032. <span class="sig-name descname"><span class="pre">normalize_targets</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.normalize_targets" title="Permalink to this definition"></a></dt>
    2033. 

  2033. <dd><p>Whether to normalize bbox coordinates by image size (default=False).</p>
    2034. 

  2034. </dd></dl>
    2035. 

  2035. 

  2036. <dl class="py attribute">
    2037. 

  2037. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.iou_thresholds">
    2038. 

  2038. <span class="sig-name descname"><span class="pre">iou_thresholds</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.iou_thresholds" title="Permalink to this definition"></a></dt>
    2039. 

  2039. <dd><p>IoU threshold to compute the mAP (default=torch.linspace(0.5, 0.95, 10)).</p>
    2040. 

  2040. </dd></dl>
    2041. 

  2041. 

  2042. <dl class="py attribute">
    2043. 

  2043. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.recall_thresholds">
    2044. 

  2044. <span class="sig-name descname"><span class="pre">recall_thresholds</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.recall_thresholds" title="Permalink to this definition"></a></dt>
    2045. 

  2045. <dd><p>Recall threshold to compute the mAP (default=torch.linspace(0, 1, 101)).</p>
    2046. 

  2046. </dd></dl>
    2047. 

  2047. 

  2048. <dl class="py attribute">
    2049. 

  2049. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.score_threshold">
    2050. 

  2050. <span class="sig-name descname"><span class="pre">score_threshold</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.score_threshold" title="Permalink to this definition"></a></dt>
    2051. 

  2051. <dd><p>Score threshold to compute Recall, Precision and F1 (default=0.1)</p>
    2052. 

  2052. </dd></dl>
    2053. 

  2053. 

  2054. <dl class="py attribute">
    2055. 

  2055. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.top_k_predictions">
    2056. 

  2056. <span class="sig-name descname"><span class="pre">top_k_predictions</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.top_k_predictions" title="Permalink to this definition"></a></dt>
    2057. 

  2057. <dd><p>Number of predictions per class used to compute metrics, ordered by confidence score
    2058. 

  2058. (default=100)</p>
    2059. 

  2059. </dd></dl>
    2060. 

  2060. 

  2061. <dl class="py attribute">
    2062. 

  2062. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.dist_sync_on_step">
    2063. 

  2063. <span class="sig-name descname"><span class="pre">dist_sync_on_step</span></span><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.dist_sync_on_step" title="Permalink to this definition"></a></dt>
    2064. 

  2064. <dd><p>Synchronize metric state across processes at each <code class="docutils literal notranslate"><span class="pre">forward()</span></code>
    2065. 

  2065. before returning the value at the step. (default=False)</p>
    2066. 

  2066. <blockquote>
    2067. 

  2067. <div><dl class="simple">
    2068. 

  2068. <dt>accumulate_on_cpu:     Run on CPU regardless of device used in other parts.</dt><dd><p>This is to avoid “CUDA out of memory” that might happen on GPU (default False)</p>
    2069. 

  2069. </dd>
    2070. 

  2070. </dl>
    2071. 

  2071. </div></blockquote>
    2072. 

  2072. </dd></dl>
    2073. 

  2073. 

  2074. <dl class="py method">
    2075. 

  2075. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.update">
    2076. 

  2076. <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">preds</span></span></em>, <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">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">inputs</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torch._VariableFunctionsClass.tensor</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">crowd_targets</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">torch.Tensor</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/metrics/detection_metrics.html#DetectionMetrics.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.update" title="Permalink to this definition"></a></dt>
    2077. 

  2077. <dd><p>Apply NMS and match all the predictions and targets of a given batch, and update the metric state accordingly.</p>
    2078. 

  2078. <dl class="simple">
    2079. 

  2079. <dt>:param preds<span class="classifier">Raw output of the model, the format might change from one model to another, but has to fit</span></dt><dd><p>the input format of the post_prediction_callback</p>
    2080. 

  2080. </dd>
    2081. 

  2081. </dl>
    2082. 

  2082. <dl class="field-list simple">
    2083. 

  2083. <dt class="field-odd">Parameters</dt>
    2084. 

  2084. <dd class="field-odd"><ul class="simple">
    2085. 

  2085. <li><p><strong>target</strong> – Targets for all images of shape (total_num_targets, 6)
    2086. 

  2086. format:  (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</p></li>
    2087. 

  2087. <li><p><strong>device</strong> – Device to run on</p></li>
    2088. 

  2088. <li><p><strong>inputs</strong> – Input image tensor of shape (batch_size, n_img, height, width)</p></li>
    2089. 

  2089. <li><p><strong>crowd_targets</strong> – Crowd targets for all images of shape (total_num_targets, 6)
    2090. 

  2090. format:  (index, x, y, w, h, label) where x,y,w,h are in range [0,1]</p></li>
    2091. 

  2091. </ul>
    2092. 

  2092. </dd>
    2093. 

  2093. </dl>
    2094. 

  2094. </dd></dl>
    2095. 

  2095. 

  2096. <dl class="py method">
    2097. 

  2097. <dt class="sig sig-object py" id="super_gradients.training.metrics.DetectionMetrics.compute">
    2098. 

  2098. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span> &#x2192; <span class="pre">Dict</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">float</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.Tensor</span><span class="p"><span class="pre">]</span></span><span class="p"><span class="pre">]</span></span><a class="reference internal" href="_modules/super_gradients/training/metrics/detection_metrics.html#DetectionMetrics.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.DetectionMetrics.compute" title="Permalink to this definition"></a></dt>
    2099. 

  2099. <dd><p>Compute the metrics for all the accumulated results.
    2100. 

  2100. :return: Metrics of interest</p>
    2101. 

  2101. </dd></dl>
    2102. 

  2102. 

  2103. </dd></dl>
    2104. 

  2104. 

  2105. <dl class="py class">
    2106. 

  2106. <dt class="sig sig-object py" id="super_gradients.training.metrics.PreprocessSegmentationMetricsArgs">
    2107. 

  2107. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">PreprocessSegmentationMetricsArgs</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">apply_arg_max</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">apply_sigmoid</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/metrics/segmentation_metrics.html#PreprocessSegmentationMetricsArgs"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.PreprocessSegmentationMetricsArgs" title="Permalink to this definition"></a></dt>
    2108. 

  2108. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></code></a></p>
    2109. 

  2109. <p>Default segmentation inputs preprocess function before updating segmentation metrics, handles multiple inputs and
    2110. 

  2110. apply normalizations.</p>
    2111. 

  2111. </dd></dl>
    2112. 

  2112. 

  2113. <dl class="py class">
    2114. 

  2114. <dt class="sig sig-object py" id="super_gradients.training.metrics.PixelAccuracy">
    2115. 

  2115. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">PixelAccuracy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ignore_label</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">-</span> <span class="pre">100</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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">metrics_args_prep_fn</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><a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></a><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/metrics/segmentation_metrics.html#PixelAccuracy"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.PixelAccuracy" title="Permalink to this definition"></a></dt>
    2116. 

  2116. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.metric.Metric</span></code></p>
    2117. 

  2117. <dl class="py method">
    2118. 

  2118. <dt class="sig sig-object py" id="super_gradients.training.metrics.PixelAccuracy.update">
    2119. 

  2119. <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">preds</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">target</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/metrics/segmentation_metrics.html#PixelAccuracy.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.PixelAccuracy.update" title="Permalink to this definition"></a></dt>
    2120. 

  2120. <dd><p>Override this method to update the state variables of your metric class.</p>
    2121. 

  2121. </dd></dl>
    2122. 

  2122. 

  2123. <dl class="py method">
    2124. 

  2124. <dt class="sig sig-object py" id="super_gradients.training.metrics.PixelAccuracy.compute">
    2125. 

  2125. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#PixelAccuracy.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.PixelAccuracy.compute" title="Permalink to this definition"></a></dt>
    2126. 

  2126. <dd><p>Override this method to compute the final metric value from state variables synchronized across the
    2127. 

  2127. distributed backend.</p>
    2128. 

  2128. </dd></dl>
    2129. 

  2129. 

  2130. </dd></dl>
    2131. 

  2131. 

  2132. <dl class="py class">
    2133. 

  2133. <dt class="sig sig-object py" id="super_gradients.training.metrics.IoU">
    2134. 

  2134. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">IoU</span></span><span class="sig-paren">(</span><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">dist_sync_on_step</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">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>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'elementwise_mean'</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">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">metrics_args_prep_fn</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><a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></a><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/metrics/segmentation_metrics.html#IoU"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.IoU" title="Permalink to this definition"></a></dt>
    2135. 

  2135. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.classification.jaccard.JaccardIndex</span></code></p>
    2136. 

  2136. <dl class="py method">
    2137. 

  2137. <dt class="sig sig-object py" id="super_gradients.training.metrics.IoU.update">
    2138. 

  2138. <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">preds</span></span></em>, <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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#IoU.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.IoU.update" title="Permalink to this definition"></a></dt>
    2139. 

  2139. <dd><p>Update state with predictions and targets.</p>
    2140. 

  2140. <dl class="field-list simple">
    2141. 

  2141. <dt class="field-odd">Parameters</dt>
    2142. 

  2142. <dd class="field-odd"><ul class="simple">
    2143. 

  2143. <li><p><strong>preds</strong> – Predictions from model</p></li>
    2144. 

  2144. <li><p><strong>target</strong> – Ground truth values</p></li>
    2145. 

  2145. </ul>
    2146. 

  2146. </dd>
    2147. 

  2147. </dl>
    2148. 

  2148. </dd></dl>
    2149. 

  2149. 

  2150. <dl class="py attribute">
    2151. 

  2151. <dt class="sig sig-object py" id="super_gradients.training.metrics.IoU.confmat">
    2152. 

  2152. <span class="sig-name descname"><span class="pre">confmat</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.IoU.confmat" title="Permalink to this definition"></a></dt>
    2153. 

  2153. <dd></dd></dl>
    2154. 

  2154. 

  2155. </dd></dl>
    2156. 

  2156. 

  2157. <dl class="py class">
    2158. 

  2158. <dt class="sig sig-object py" id="super_gradients.training.metrics.Dice">
    2159. 

  2159. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">Dice</span></span><span class="sig-paren">(</span><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">dist_sync_on_step</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">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>, <em class="sig-param"><span class="n"><span class="pre">reduction</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">str</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">'elementwise_mean'</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">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">metrics_args_prep_fn</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><a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></a><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/metrics/segmentation_metrics.html#Dice"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Dice" title="Permalink to this definition"></a></dt>
    2160. 

  2160. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torchmetrics.classification.jaccard.JaccardIndex</span></code></p>
    2161. 

  2161. <dl class="py method">
    2162. 

  2162. <dt class="sig sig-object py" id="super_gradients.training.metrics.Dice.update">
    2163. 

  2163. <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">preds</span></span></em>, <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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#Dice.update"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Dice.update" title="Permalink to this definition"></a></dt>
    2164. 

  2164. <dd><p>Update state with predictions and targets.</p>
    2165. 

  2165. <dl class="field-list simple">
    2166. 

  2166. <dt class="field-odd">Parameters</dt>
    2167. 

  2167. <dd class="field-odd"><ul class="simple">
    2168. 

  2168. <li><p><strong>preds</strong> – Predictions from model</p></li>
    2169. 

  2169. <li><p><strong>target</strong> – Ground truth values</p></li>
    2170. 

  2170. </ul>
    2171. 

  2171. </dd>
    2172. 

  2172. </dl>
    2173. 

  2173. </dd></dl>
    2174. 

  2174. 

  2175. <dl class="py method">
    2176. 

  2176. <dt class="sig sig-object py" id="super_gradients.training.metrics.Dice.compute">
    2177. 

  2177. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span> &#x2192; <span class="pre">torch.Tensor</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#Dice.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.Dice.compute" title="Permalink to this definition"></a></dt>
    2178. 

  2178. <dd><p>Computes Dice coefficient</p>
    2179. 

  2179. </dd></dl>
    2180. 

  2180. 

  2181. <dl class="py attribute">
    2182. 

  2182. <dt class="sig sig-object py" id="super_gradients.training.metrics.Dice.confmat">
    2183. 

  2183. <span class="sig-name descname"><span class="pre">confmat</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.Dice.confmat" title="Permalink to this definition"></a></dt>
    2184. 

  2184. <dd></dd></dl>
    2185. 

  2185. 

  2186. </dd></dl>
    2187. 

  2187. 

  2188. <dl class="py class">
    2189. 

  2189. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryIOU">
    2190. 

  2190. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">BinaryIOU</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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">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>, <em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metrics_args_prep_fn</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><a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></a><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/metrics/segmentation_metrics.html#BinaryIOU"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.BinaryIOU" title="Permalink to this definition"></a></dt>
    2191. 

  2191. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.IoU" title="super_gradients.training.metrics.segmentation_metrics.IoU"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.metrics.segmentation_metrics.IoU</span></code></a></p>
    2192. 

  2192. <dl class="py method">
    2193. 

  2193. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryIOU.compute">
    2194. 

  2194. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#BinaryIOU.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.BinaryIOU.compute" title="Permalink to this definition"></a></dt>
    2195. 

  2195. <dd><p>Computes intersection over union (IoU)</p>
    2196. 

  2196. </dd></dl>
    2197. 

  2197. 

  2198. <dl class="py attribute">
    2199. 

  2199. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryIOU.confmat">
    2200. 

  2200. <span class="sig-name descname"><span class="pre">confmat</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.BinaryIOU.confmat" title="Permalink to this definition"></a></dt>
    2201. 

  2201. <dd></dd></dl>
    2202. 

  2202. 

  2203. <dl class="py attribute">
    2204. 

  2204. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryIOU.training">
    2205. 

  2205. <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.metrics.BinaryIOU.training" title="Permalink to this definition"></a></dt>
    2206. 

  2206. <dd></dd></dl>
    2207. 

  2207. 

  2208. </dd></dl>
    2209. 

  2209. 

  2210. <dl class="py class">
    2211. 

  2211. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryDice">
    2212. 

  2212. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.metrics.</span></span><span class="sig-name descname"><span class="pre">BinaryDice</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dist_sync_on_step</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">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>, <em class="sig-param"><span class="n"><span class="pre">threshold</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">float</span></span> <span class="o"><span class="pre">=</span></span> <span class="default_value"><span class="pre">0.5</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metrics_args_prep_fn</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><a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn" title="super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn"><span class="pre">super_gradients.training.metrics.segmentation_metrics.AbstractMetricsArgsPrepFn</span></a><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/metrics/segmentation_metrics.html#BinaryDice"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.BinaryDice" title="Permalink to this definition"></a></dt>
    2213. 

  2213. <dd><p>Bases: <a class="reference internal" href="super_gradients.training.metrics.html#super_gradients.training.metrics.segmentation_metrics.Dice" title="super_gradients.training.metrics.segmentation_metrics.Dice"><code class="xref py py-class docutils literal notranslate"><span class="pre">super_gradients.training.metrics.segmentation_metrics.Dice</span></code></a></p>
    2214. 

  2214. <dl class="py method">
    2215. 

  2215. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryDice.compute">
    2216. 

  2216. <span class="sig-name descname"><span class="pre">compute</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/metrics/segmentation_metrics.html#BinaryDice.compute"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.metrics.BinaryDice.compute" title="Permalink to this definition"></a></dt>
    2217. 

  2217. <dd><p>Computes Dice coefficient</p>
    2218. 

  2218. </dd></dl>
    2219. 

  2219. 

  2220. <dl class="py attribute">
    2221. 

  2221. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryDice.confmat">
    2222. 

  2222. <span class="sig-name descname"><span class="pre">confmat</span></span><em class="property"><span class="pre">:</span> <span class="pre">torch.Tensor</span></em><a class="headerlink" href="#super_gradients.training.metrics.BinaryDice.confmat" title="Permalink to this definition"></a></dt>
    2223. 

  2223. <dd></dd></dl>
    2224. 

  2224. 

  2225. <dl class="py attribute">
    2226. 

  2226. <dt class="sig sig-object py" id="super_gradients.training.metrics.BinaryDice.training">
    2227. 

  2227. <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.metrics.BinaryDice.training" title="Permalink to this definition"></a></dt>
    2228. 

  2228. <dd></dd></dl>
    2229. 

  2229. 

  2230. </dd></dl>
    2231. 

  2231. 

  2232. </section>
    2233. 

  2233. <section id="module-super_gradients.training.models">
    2234. 

  2234. <span id="super-gradients-training-models-module"></span><h2>super_gradients.training.models module<a class="headerlink" href="#module-super_gradients.training.models" title="Permalink to this headline"></a></h2>
    2235. 

  2235. </section>
    2236. 

  2236. <section id="module-super_gradients.training.sg_model">
    2237. 

  2237. <span id="super-gradients-training-sg-model-module"></span><h2>super_gradients.training.sg_model module<a class="headerlink" href="#module-super_gradients.training.sg_model" title="Permalink to this headline"></a></h2>
    2238. 

  2238. <dl class="py class">
    2239. 

  2239. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel">
    2240. 

  2240. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.sg_model.</span></span><span class="sig-name descname"><span class="pre">SgModel</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="pre">experiment_name:</span> <span class="pre">str</span></em>, <em class="sig-param"><span class="pre">device:</span> <span class="pre">Optional[str]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">multi_gpu:</span> <span class="pre">Union[super_gradients.common.data_types.enum.multi_gpu_mode.MultiGPUMode</span></em>, <em class="sig-param"><span class="pre">str]</span> <span class="pre">=</span> <span class="pre">&lt;MultiGPUMode.OFF:</span> <span class="pre">'Off'&gt;</span></em>, <em class="sig-param"><span class="pre">model_checkpoints_location:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'local'</span></em>, <em class="sig-param"><span class="pre">overwrite_local_checkpoint:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></em>, <em class="sig-param"><span class="pre">ckpt_name:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'ckpt_latest.pth'</span></em>, <em class="sig-param"><span class="pre">post_prediction_callback:</span> <span class="pre">Optional[super_gradients.training.utils.detection_utils.DetectionPostPredictionCallback]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">ckpt_root_dir:</span> <span class="pre">Optional[str]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">train_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">valid_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">test_loader:</span> <span class="pre">Optional[torch.utils.data.dataloader.DataLoader]</span> <span class="pre">=</span> <span class="pre">None</span></em>, <em class="sig-param"><span class="pre">classes:</span> <span class="pre">Optional[List[Any]]</span> <span class="pre">=</span> <span class="pre">None</span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel" title="Permalink to this definition"></a></dt>
    2241. 

  2241. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    2242. 

  2242. <p>SuperGradient Model - Base Class for Sg Models</p>
    2243. 

  2243. <dl class="py method">
    2244. 

  2244. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.train">
    2245. 

  2245. <span class="sig-name descname"><span class="pre">train</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">max_epochs</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">initial_epoch</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">save_model</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/sg_model/sg_model.html#SgModel.train"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.train" title="Permalink to this definition"></a></dt>
    2246. 

  2246. <dd><p>the main function used for the training, h.p. updating, logging etc.</p>
    2247. 

  2247. </dd></dl>
    2248. 

  2248. 

  2249. <dl class="py method">
    2250. 

  2250. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.predict">
    2251. 

  2251. <span class="sig-name descname"><span class="pre">predict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">idx</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/sg_model/sg_model.html#SgModel.predict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.predict" title="Permalink to this definition"></a></dt>
    2252. 

  2252. <dd><p>returns the predictions and label of the current inputs</p>
    2253. 

  2253. </dd></dl>
    2254. 

  2254. 

  2255. <dl class="py method">
    2256. 

  2256. <dt class="sig sig-object py">
    2257. 

  2257. <span class="sig-name descname"><span class="pre">test(epoch</span> <span class="pre">:</span> <span class="pre">int,</span> <span class="pre">idx</span> <span class="pre">:</span> <span class="pre">int,</span> <span class="pre">save</span> <span class="pre">:</span> <span class="pre">bool):</span></span></dt>
    2258. 

  2258. <dd><p>returns the test loss, accuracy and runtime</p>
    2259. 

  2259. </dd></dl>
    2260. 

  2260. 

  2261. <dl class="py method">
    2262. 

  2262. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.connect_dataset_interface">
    2263. 

  2263. <span class="sig-name descname"><span class="pre">connect_dataset_interface</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dataset_interface</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.training.datasets.dataset_interfaces.html#super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface" title="super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface"><span class="pre">super_gradients.training.datasets.dataset_interfaces.dataset_interface.DatasetInterface</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">data_loader_num_workers</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">8</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.connect_dataset_interface"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.connect_dataset_interface" title="Permalink to this definition"></a></dt>
    2264. 

  2264. <dd><dl class="field-list simple">
    2265. 

  2265. <dt class="field-odd">Parameters</dt>
    2266. 

  2266. <dd class="field-odd"><ul class="simple">
    2267. 

  2267. <li><p><strong>dataset_interface</strong> – DatasetInterface object</p></li>
    2268. 

  2268. <li><p><strong>data_loader_num_workers</strong> – The number of threads to initialize the Data Loaders with
    2269. 

  2269. The dataset to be connected</p></li>
    2270. 

  2270. </ul>
    2271. 

  2271. </dd>
    2272. 

  2272. </dl>
    2273. 

  2273. </dd></dl>
    2274. 

  2274. 

  2275. <dl class="py method">
    2276. 

  2276. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.build_model">
    2277. 

  2277. <span class="sig-name descname"><span class="pre">build_model</span></span><span class="sig-paren">(</span><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">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">str</span><span class="p"><span class="pre">,</span> </span><span class="pre">torch.nn.modules.module.Module</span><span class="p"><span class="pre">]</span></span></span></em>, <em class="sig-param"><span class="n"><span class="pre">arch_params</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">checkpoint_params</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="o"><span class="pre">*</span></span><span class="n"><span class="pre">args</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/sg_model/sg_model.html#SgModel.build_model"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.build_model" title="Permalink to this definition"></a></dt>
    2278. 

  2278. <dd><dl class="field-list simple">
    2279. 

  2279. <dt class="field-odd">Parameters</dt>
    2280. 

  2280. <dd class="field-odd"><ul class="simple">
    2281. 

  2281. <li><p><strong>architecture</strong> – Defines the network’s architecture from models/ALL_ARCHITECTURES</p></li>
    2282. 

  2282. <li><p><strong>arch_params</strong> – Architecture H.P. e.g.: block, num_blocks, num_classes, etc.</p></li>
    2283. 

  2283. <li><p><strong>checkpoint_params</strong> – <p>Dictionary like object with the following key:values:</p>
    2284. 

  2284. <p>load_checkpoint:            Load a pre-trained checkpoint
    2285. 

  2285. strict_load:                See StrictLoad class documentation for details.
    2286. 

  2286. source_ckpt_folder_name:    folder name to load the checkpoint from (self.experiment_name if none is given)
    2287. 

  2287. load_weights_only:          loads only the weight from the checkpoint and zeroize the training params
    2288. 

  2288. load_backbone:              loads the provided checkpoint to self.net.backbone instead of self.net
    2289. 

  2289. external_checkpoint_path:   The path to the external checkpoint to be loaded. Can be absolute or relative</p>
    2290. 

  2290. <blockquote>
    2291. 

  2291. <div><p>(ie: path/to/checkpoint.pth). If provided, will automatically attempt to
    2292. 

  2292. load the checkpoint even if the load_checkpoint flag is not provided.</p>
    2293. 

  2293. </div></blockquote>
    2294. 

  2294. </p></li>
    2295. 

  2295. </ul>
    2296. 

  2296. </dd>
    2297. 

  2297. </dl>
    2298. 

  2298. </dd></dl>
    2299. 

  2299. 

  2300. <dl class="py method">
    2301. 

  2301. <dt class="sig sig-object py" id="id4">
    2302. 

  2302. <span class="sig-name descname"><span class="pre">train</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">training_params</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">dict</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/sg_model/sg_model.html#SgModel.train"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#id4" title="Permalink to this definition"></a></dt>
    2303. 

  2303. <dd><p>train - Trains the Model</p>
    2304. 

  2304. <dl>
    2305. 

  2305. <dt>IMPORTANT NOTE: Additional batch parameters can be added as a third item (optional) if a tuple is returned by</dt><dd><p>the data loaders, as dictionary. The phase context will hold the additional items, under an attribute with
    2306. 

  2306. the same name as the key in this dictionary. Then such items can be accessed through phase callbacks.</p>
    2307. 

  2307. <blockquote>
    2308. 

  2308. <div><dl class="field-list">
    2309. 

  2309. <dt class="field-odd">param training_params</dt>
    2310. 

  2310. <dd class="field-odd"><ul>
    2311. 

  2311. <li><p><cite>max_epochs</cite> : int</p>
    2312. 

  2312. <blockquote>
    2313. 

  2313. <div><p>Number of epochs to run training.</p>
    2314. 

  2314. </div></blockquote>
    2315. 

  2315. </li>
    2316. 

  2316. <li><p><cite>lr_updates</cite> : list(int)</p>
    2317. 

  2317. <blockquote>
    2318. 

  2318. <div><p>List of fixed epoch numbers to perform learning rate updates when <cite>lr_mode=’step’</cite>.</p>
    2319. 

  2319. </div></blockquote>
    2320. 

  2320. </li>
    2321. 

  2321. <li><p><cite>lr_decay_factor</cite> : float</p>
    2322. 

  2322. <blockquote>
    2323. 

  2323. <div><p>Decay factor to apply to the learning rate at each update when <cite>lr_mode=’step’</cite>.</p>
    2324. 

  2324. </div></blockquote>
    2325. 

  2325. </li>
    2326. 

  2326. <li><p><cite>lr_mode</cite> : str</p>
    2327. 

  2327. <blockquote>
    2328. 

  2328. <div><p>Learning rate scheduling policy, one of [‘step’,’poly’,’cosine’,’function’]. ‘step’ refers to
    2329. 

  2329. constant updates at epoch numbers passed through <cite>lr_updates</cite>. ‘cosine’ refers to Cosine Anealing
    2330. 

  2330. policy as mentioned in <a class="reference external" href="https://arxiv.org/abs/1608.03983">https://arxiv.org/abs/1608.03983</a>. ‘poly’ refers to polynomial decrease i.e
    2331. 

  2331. in each epoch iteration <cite>self.lr = self.initial_lr * pow((1.0 - (current_iter / max_iter)),
    2332. 

  2332. 0.9)</cite> ‘function’ refers to user defined learning rate scheduling function, that is passed through
    2333. 

  2333. <cite>lr_schedule_function</cite>.</p>
    2334. 

  2334. </div></blockquote>
    2335. 

  2335. </li>
    2336. 

  2336. <li><p><cite>lr_schedule_function</cite> : Union[callable,None]</p>
    2337. 

  2337. <blockquote>
    2338. 

  2338. <div><p>Learning rate scheduling function to be used when <cite>lr_mode</cite> is ‘function’.</p>
    2339. 

  2339. </div></blockquote>
    2340. 

  2340. </li>
    2341. 

  2341. <li><p><cite>lr_warmup_epochs</cite> : int (default=0)</p>
    2342. 

  2342. <blockquote>
    2343. 

  2343. <div><p>Number of epochs for learning rate warm up - see <a class="reference external" href="https://arxiv.org/pdf/1706.02677.pdf">https://arxiv.org/pdf/1706.02677.pdf</a> (Section 2.2).</p>
    2344. 

  2344. </div></blockquote>
    2345. 

  2345. </li>
    2346. 

  2346. <li><dl class="simple">
    2347. 

  2347. <dt><cite>cosine_final_lr_ratio</cite><span class="classifier">float (default=0.01)</span></dt><dd><dl class="simple">
    2348. 

  2348. <dt>Final learning rate ratio (only relevant when <a href="#id5"><span class="problematic" id="id6">`</span></a>lr_mode`=’cosine’). The cosine starts from initial_lr and reaches</dt><dd><p>initial_lr * cosine_final_lr_ratio in last epoch</p>
    2349. 

  2349. </dd>
    2350. 

  2350. </dl>
    2351. 

  2351. </dd>
    2352. 

  2352. </dl>
    2353. 

  2353. </li>
    2354. 

  2354. <li><p><cite>inital_lr</cite> : float</p>
    2355. 

  2355. <blockquote>
    2356. 

  2356. <div><p>Initial learning rate.</p>
    2357. 

  2357. </div></blockquote>
    2358. 

  2358. </li>
    2359. 

  2359. <li><p><cite>loss</cite> : Union[nn.module, str]</p>
    2360. 

  2360. <blockquote>
    2361. 

  2361. <div><p>Loss function for training.
    2362. 

  2362. One of SuperGradient’s built in options:</p>
    2363. 

  2363. <blockquote>
    2364. 

  2364. <div><p>“cross_entropy”: LabelSmoothingCrossEntropyLoss,
    2365. 

  2365. “mse”: MSELoss,
    2366. 

  2366. “r_squared_loss”: RSquaredLoss,
    2367. 

  2367. “detection_loss”: YoLoV3DetectionLoss,
    2368. 

  2368. “shelfnet_ohem_loss”: ShelfNetOHEMLoss,
    2369. 

  2369. “shelfnet_se_loss”: ShelfNetSemanticEncodingLoss,
    2370. 

  2370. “ssd_loss”: SSDLoss,</p>
    2371. 

  2371. </div></blockquote>
    2372. 

  2372. <p>or user defined nn.module loss function.</p>
    2373. 

  2373. <p>IMPORTANT: forward(…) should return a (loss, loss_items) tuple where loss is the tensor used
    2374. 

  2374. for backprop (i.e what your original loss function returns), and loss_items should be a tensor of
    2375. 

  2375. shape (n_items), of values computed during the forward pass which we desire to log over the
    2376. 

  2376. entire epoch. For example- the loss itself should always be logged. Another example is a scenario
    2377. 

  2377. where the computed loss is the sum of a few components we would like to log- these entries in
    2378. 

  2378. loss_items).</p>
    2379. 

  2379. <p>When training, set the loss_logging_items_names parameter in train_params to be a list of
    2380. 

  2380. strings, of length n_items who’s ith element is the name of the ith entry in loss_items. Then
    2381. 

  2381. each item will be logged, rendered on tensorboard and “watched” (i.e saving model checkpoints
    2382. 

  2382. according to it).</p>
    2383. 

  2383. <p>Since running logs will save the loss_items in some internal state, it is recommended that
    2384. 

  2384. loss_items are detached from their computational graph for memory efficiency.</p>
    2385. 

  2385. </div></blockquote>
    2386. 

  2386. </li>
    2387. 

  2387. <li><p><cite>optimizer</cite> : Union[str, torch.optim.Optimizer]</p>
    2388. 

  2388. <blockquote>
    2389. 

  2389. <div><p>Optimization algorithm. One of [‘Adam’,’SGD’,’RMSProp’] corresponding to the torch.optim
    2390. 

  2390. optimzers implementations, or any object that implements torch.optim.Optimizer.</p>
    2391. 

  2391. </div></blockquote>
    2392. 

  2392. </li>
    2393. 

  2393. <li><p><cite>criterion_params</cite> : dict</p>
    2394. 

  2394. <blockquote>
    2395. 

  2395. <div><p>Loss function parameters.</p>
    2396. 

  2396. </div></blockquote>
    2397. 

  2397. </li>
    2398. 

  2398. <li><dl>
    2399. 

  2399. <dt><cite>optimizer_params</cite><span class="classifier">dict</span></dt><dd><p>When <cite>optimizer</cite> is one of [‘Adam’,’SGD’,’RMSProp’], it will be initialized with optimizer_params.</p>
    2400. 

  2400. <p>(see <a class="reference external" href="https://pytorch.org/docs/stable/optim.html">https://pytorch.org/docs/stable/optim.html</a> for the full list of
    2401. 

  2401. parameters for each optimizer).</p>
    2402. 

  2402. </dd>
    2403. 

  2403. </dl>
    2404. 

  2404. </li>
    2405. 

  2405. <li><p><cite>train_metrics_list</cite> : list(torchmetrics.Metric)</p>
    2406. 

  2406. <blockquote>
    2407. 

  2407. <div><p>Metrics to log during training. For more information on torchmetrics see
    2408. 

  2408. <a class="reference external" href="https://torchmetrics.rtfd.io/en/latest/">https://torchmetrics.rtfd.io/en/latest/</a>.</p>
    2409. 

  2409. </div></blockquote>
    2410. 

  2410. </li>
    2411. 

  2411. <li><p><cite>valid_metrics_list</cite> : list(torchmetrics.Metric)</p>
    2412. 

  2412. <blockquote>
    2413. 

  2413. <div><p>Metrics to log during validation/testing. For more information on torchmetrics see
    2414. 

  2414. <a class="reference external" href="https://torchmetrics.rtfd.io/en/latest/">https://torchmetrics.rtfd.io/en/latest/</a>.</p>
    2415. 

  2415. </div></blockquote>
    2416. 

  2416. </li>
    2417. 

  2417. <li><p><cite>loss_logging_items_names</cite> : list(str)</p>
    2418. 

  2418. <blockquote>
    2419. 

  2419. <div><p>The list of names/titles for the outputs returned from the loss functions forward pass (reminder-
    2420. 

  2420. the loss function should return the tuple (loss, loss_items)). These names will be used for
    2421. 

  2421. logging their values.</p>
    2422. 

  2422. </div></blockquote>
    2423. 

  2423. </li>
    2424. 

  2424. <li><p><cite>metric_to_watch</cite> : str (default=”Accuracy”)</p>
    2425. 

  2425. <blockquote>
    2426. 

  2426. <div><p>will be the metric which the model checkpoint will be saved according to, and can be set to any
    2427. 

  2427. of the following:</p>
    2428. 

  2428. <blockquote>
    2429. 

  2429. <div><p>a metric name (str) of one of the metric objects from the valid_metrics_list</p>
    2430. 

  2430. <p>a “metric_name” if some metric in valid_metrics_list has an attribute component_names which
    2431. 

  2431. is a list referring to the names of each entry in the output metric (torch tensor of size n)</p>
    2432. 

  2432. <p>one of “loss_logging_items_names” i.e which will correspond to an item returned during the
    2433. 

  2433. loss function’s forward pass.</p>
    2434. 

  2434. </div></blockquote>
    2435. 

  2435. <p>At the end of each epoch, if a new best metric_to_watch value is achieved, the models checkpoint
    2436. 

  2436. is saved in YOUR_PYTHON_PATH/checkpoints/ckpt_best.pth</p>
    2437. 

  2437. </div></blockquote>
    2438. 

  2438. </li>
    2439. 

  2439. <li><p><cite>greater_metric_to_watch_is_better</cite> : bool</p>
    2440. 

  2440. <blockquote>
    2441. 

  2441. <div><dl class="simple">
    2442. 

  2442. <dt>When choosing a model’s checkpoint to be saved, the best achieved model is the one that maximizes the</dt><dd><p>metric_to_watch when this parameter is set to True, and a one that minimizes it otherwise.</p>
    2443. 

  2443. </dd>
    2444. 

  2444. </dl>
    2445. 

  2445. </div></blockquote>
    2446. 

  2446. </li>
    2447. 

  2447. <li><p><cite>ema</cite> : bool (default=False)</p>
    2448. 

  2448. <blockquote>
    2449. 

  2449. <div><p>Whether to use Model Exponential Moving Average (see
    2450. 

  2450. <a class="reference external" href="https://github.com/rwightman/pytorch-image-models">https://github.com/rwightman/pytorch-image-models</a> ema implementation)</p>
    2451. 

  2451. </div></blockquote>
    2452. 

  2452. </li>
    2453. 

  2453. <li><p><cite>batch_accumulate</cite> : int (default=1)</p>
    2454. 

  2454. <blockquote>
    2455. 

  2455. <div><p>Number of batches to accumulate before every backward pass.</p>
    2456. 

  2456. </div></blockquote>
    2457. 

  2457. </li>
    2458. 

  2458. <li><p><cite>ema_params</cite> : dict</p>
    2459. 

  2459. <blockquote>
    2460. 

  2460. <div><p>Parameters for the ema model.</p>
    2461. 

  2461. </div></blockquote>
    2462. 

  2462. </li>
    2463. 

  2463. <li><p><cite>zero_weight_decay_on_bias_and_bn</cite> : bool (default=False)</p>
    2464. 

  2464. <blockquote>
    2465. 

  2465. <div><p>Whether to apply weight decay on batch normalization parameters or not (ignored when the passed
    2466. 

  2466. optimizer has already been initialized).</p>
    2467. 

  2467. </div></blockquote>
    2468. 

  2468. </li>
    2469. 

  2469. <li><p><cite>load_opt_params</cite> : bool (default=True)</p>
    2470. 

  2470. <blockquote>
    2471. 

  2471. <div><p>Whether to load the optimizers parameters as well when loading a model’s checkpoint.</p>
    2472. 

  2472. </div></blockquote>
    2473. 

  2473. </li>
    2474. 

  2474. <li><p><cite>run_validation_freq</cite> : int (default=1)</p>
    2475. 

  2475. <blockquote>
    2476. 

  2476. <div><dl class="simple">
    2477. 

  2477. <dt>The frequency in which validation is performed during training (i.e the validation is ran every</dt><dd><p><cite>run_validation_freq</cite> epochs.</p>
    2478. 

  2478. </dd>
    2479. 

  2479. </dl>
    2480. 

  2480. </div></blockquote>
    2481. 

  2481. </li>
    2482. 

  2482. <li><p><cite>save_model</cite> : bool (default=True)</p>
    2483. 

  2483. <blockquote>
    2484. 

  2484. <div><p>Whether to save the model checkpoints.</p>
    2485. 

  2485. </div></blockquote>
    2486. 

  2486. </li>
    2487. 

  2487. <li><p><cite>silent_mode</cite> : bool</p>
    2488. 

  2488. <blockquote>
    2489. 

  2489. <div><p>Silents the print outs.</p>
    2490. 

  2490. </div></blockquote>
    2491. 

  2491. </li>
    2492. 

  2492. <li><p><cite>mixed_precision</cite> : bool</p>
    2493. 

  2493. <blockquote>
    2494. 

  2494. <div><p>Whether to use mixed precision or not.</p>
    2495. 

  2495. </div></blockquote>
    2496. 

  2496. </li>
    2497. 

  2497. <li><p><cite>save_ckpt_epoch_list</cite> : list(int) (default=[])</p>
    2498. 

  2498. <blockquote>
    2499. 

  2499. <div><p>List of fixed epoch indices the user wishes to save checkpoints in.</p>
    2500. 

  2500. </div></blockquote>
    2501. 

  2501. </li>
    2502. 

  2502. <li><p><cite>average_best_models</cite> : bool (default=False)</p>
    2503. 

  2503. <blockquote>
    2504. 

  2504. <div><p>If set, a snapshot dictionary file and the average model will be saved / updated at every epoch
    2505. 

  2505. and evaluated only when training is completed. The snapshot file will only be deleted upon
    2506. 

  2506. completing the training. The snapshot dict will be managed on cpu.</p>
    2507. 

  2507. </div></blockquote>
    2508. 

  2508. </li>
    2509. 

  2509. <li><p><cite>precise_bn</cite> : bool (default=False)</p>
    2510. 

  2510. <blockquote>
    2511. 

  2511. <div><p>Whether to use precise_bn calculation during the training.</p>
    2512. 

  2512. </div></blockquote>
    2513. 

  2513. </li>
    2514. 

  2514. <li><p><cite>precise_bn_batch_size</cite> : int (default=None)</p>
    2515. 

  2515. <blockquote>
    2516. 

  2516. <div><p>The effective batch size we want to calculate the batchnorm on. For example, if we are training a model
    2517. 

  2517. on 8 gpus, with a batch of 128 on each gpu, a good rule of thumb would be to give it 8192
    2518. 

  2518. (ie: effective_batch_size * num_gpus = batch_per_gpu * num_gpus * num_gpus).
    2519. 

  2519. If precise_bn_batch_size is not provided in the training_params, the latter heuristic will be taken.</p>
    2520. 

  2520. </div></blockquote>
    2521. 

  2521. </li>
    2522. 

  2522. <li><p><cite>seed</cite> : int (default=42)</p>
    2523. 

  2523. <blockquote>
    2524. 

  2524. <div><p>Random seed to be set for torch, numpy, and random. When using DDP each process will have it’s seed
    2525. 

  2525. set to seed + rank.</p>
    2526. 

  2526. </div></blockquote>
    2527. 

  2527. </li>
    2528. 

  2528. <li><p><cite>log_installed_packages</cite> : bool (default=False)</p>
    2529. 

  2529. <blockquote>
    2530. 

  2530. <div><dl class="simple">
    2531. 

  2531. <dt>When set, the list of all installed packages (and their versions) will be written to the tensorboard</dt><dd><p>and logfile (useful when trying to reproduce results).</p>
    2532. 

  2532. </dd>
    2533. 

  2533. </dl>
    2534. 

  2534. </div></blockquote>
    2535. 

  2535. </li>
    2536. 

  2536. <li><p><cite>dataset_statistics</cite> : bool (default=False)</p>
    2537. 

  2537. <blockquote>
    2538. 

  2538. <div><p>Enable a statistic analysis of the dataset. If set to True the dataset will be analyzed and a report
    2539. 

  2539. will be added to the tensorboard along with some sample images from the dataset. Currently only
    2540. 

  2540. detection datasets are supported for analysis.</p>
    2541. 

  2541. </div></blockquote>
    2542. 

  2542. </li>
    2543. 

  2543. <li><p><cite>save_full_train_log</cite> : bool (default=False)</p>
    2544. 

  2544. <blockquote>
    2545. 

  2545. <div><dl class="simple">
    2546. 

  2546. <dt>When set, a full log (of all super_gradients modules, including uncaught exceptions from any other</dt><dd><p>module) of the training will be saved in the checkpoint directory under full_train_log.log</p>
    2547. 

  2547. </dd>
    2548. 

  2548. </dl>
    2549. 

  2549. </div></blockquote>
    2550. 

  2550. </li>
    2551. 

  2551. <li><p><cite>sg_logger</cite> : Union[AbstractSGLogger, str] (defauls=base_sg_logger)</p>
    2552. 

  2552. <blockquote>
    2553. 

  2553. <div><p>Define the SGLogger object for this training process. The SGLogger handles all disk writes, logs, TensorBoard, remote logging
    2554. 

  2554. and remote storage. By overriding the default base_sg_logger, you can change the storage location, support external monitoring and logging
    2555. 

  2555. or support remote storage.</p>
    2556. 

  2556. </div></blockquote>
    2557. 

  2557. </li>
    2558. 

  2558. <li><p><cite>sg_logger_params</cite> : dict</p>
    2559. 

  2559. <p>SGLogger parameters</p>
    2560. 

  2560. </li>
    2561. 

  2561. <li><p><cite>clip_grad_norm</cite> : float</p>
    2562. 

  2562. <p>Defines a maximal L2 norm of the gradients. Values which exceed the given value will be clipped</p>
    2563. 

  2563. </li>
    2564. 

  2564. <li><p><cite>lr_cooldown_epochs</cite> : int (default=0)</p>
    2565. 

  2565. <p>Number of epochs to cooldown LR (i.e the last epoch from scheduling view point=max_epochs-cooldown).</p>
    2566. 

  2566. </li>
    2567. 

  2567. <li><p><cite>pre_prediction_callback</cite> : Callable (default=None)</p>
    2568. 

  2568. <blockquote>
    2569. 

  2569. <div><dl class="simple">
    2570. 

  2570. <dt>When not None, this callback will be applied to images and targets, and returning them to be used</dt><dd><p>for the forward pass, and further computations. Args for this callable should be in the order
    2571. 

  2571. (inputs, targets, batch_idx) returning modified_inputs, modified_targets</p>
    2572. 

  2572. </dd>
    2573. 

  2573. </dl>
    2574. 

  2574. </div></blockquote>
    2575. 

  2575. </li>
    2576. 

  2576. <li><p><cite>ckpt_best_name</cite> : str (default=’ckpt_best.pth’)</p>
    2577. 

  2577. <p>The best checkpoint (according to metric_to_watch) will be saved under this filename in the checkpoints directory.</p>
    2578. 

  2578. </li>
    2579. 

  2579. <li><p><cite>enable_qat</cite>: bool (default=False)</p>
    2580. 

  2580. <dl class="simple">
    2581. 

  2581. <dt>Adds a QATCallback to the phase callbacks, that triggers quantization aware training starting from</dt><dd><p>qat_params[“start_epoch”]</p>
    2582. 

  2582. </dd>
    2583. 

  2583. </dl>
    2584. 

  2584. </li>
    2585. 

  2585. <li><p><cite>qat_params</cite>: dict-like object with the following key/values:</p>
    2586. 

  2586. <blockquote>
    2587. 

  2587. <div><p>start_epoch: int, first epoch to start QAT.</p>
    2588. 

  2588. <dl class="simple">
    2589. 

  2589. <dt>quant_modules_calib_method: str, One of [percentile, mse, entropy, max]. Statistics method for amax</dt><dd><p>computation of the quantized modules (default=percentile).</p>
    2590. 

  2590. </dd>
    2591. 

  2591. </dl>
    2592. 

  2592. <p>per_channel_quant_modules: bool, whether quant modules should be per channel (default=False).</p>
    2593. 

  2593. <p>calibrate: bool, whether to perfrom calibration (default=False).</p>
    2594. 

  2594. <p>calibrated_model_path: str, path to a calibrated checkpoint (default=None).</p>
    2595. 

  2595. <dl class="simple">
    2596. 

  2596. <dt>calib_data_loader: torch.utils.data.DataLoader, data loader of the calibration dataset. When None,</dt><dd><p>context.train_loader will be used (default=None).</p>
    2597. 

  2597. </dd>
    2598. 

  2598. </dl>
    2599. 

  2599. <p>num_calib_batches: int, number of batches to collect the statistics from.</p>
    2600. 

  2600. <dl class="simple">
    2601. 

  2601. <dt>percentile: float, percentile value to use when SgModel,quant_modules_calib_method=’percentile’.</dt><dd><p>Discarded when other methods are used (Default=99.99).</p>
    2602. 

  2602. </dd>
    2603. 

  2603. </dl>
    2604. 

  2604. </div></blockquote>
    2605. 

  2605. </li>
    2606. 

  2606. </ul>
    2607. 

  2607. </dd>
    2608. 

  2608. </dl>
    2609. 

  2609. </div></blockquote>
    2610. 

  2610. </dd>
    2611. 

  2611. </dl>
    2612. 

  2612. <dl class="field-list simple">
    2613. 

  2613. <dt class="field-odd">Returns</dt>
    2614. 

  2614. <dd class="field-odd"><p></p>
    2615. 

  2615. </dd>
    2616. 

  2616. </dl>
    2617. 

  2617. </dd></dl>
    2618. 

  2618. 

  2619. <dl class="py method">
    2620. 

  2620. <dt class="sig sig-object py" id="id7">
    2621. 

  2621. <span class="sig-name descname"><span class="pre">predict</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">inputs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">targets</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">half</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">normalize</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">verbose</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">move_outputs_to_cpu</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/sg_model/sg_model.html#SgModel.predict"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#id7" title="Permalink to this definition"></a></dt>
    2622. 

  2622. <dd><p>A fast predictor for a batch of inputs
    2623. 

  2623. :param inputs: torch.tensor or numpy.array</p>
    2624. 

  2624. <blockquote>
    2625. 

  2625. <div><p>a batch of inputs</p>
    2626. 

  2626. </div></blockquote>
    2627. 

  2627. <dl class="field-list simple">
    2628. 

  2628. <dt class="field-odd">Parameters</dt>
    2629. 

  2629. <dd class="field-odd"><ul class="simple">
    2630. 

  2630. <li><p><strong>targets</strong> – torch.tensor()
    2631. 

  2631. corresponding labels - if non are given - accuracy will not be computed</p></li>
    2632. 

  2632. <li><p><strong>verbose</strong> – bool
    2633. 

  2633. print the results to screen</p></li>
    2634. 

  2634. <li><p><strong>normalize</strong> – bool
    2635. 

  2635. If true, normalizes the tensor according to the dataloader’s normalization values</p></li>
    2636. 

  2636. <li><p><strong>half</strong> – Performs half precision evaluation</p></li>
    2637. 

  2637. <li><p><strong>move_outputs_to_cpu</strong> – Moves the results from the GPU to the CPU</p></li>
    2638. 

  2638. </ul>
    2639. 

  2639. </dd>
    2640. 

  2640. <dt class="field-even">Returns</dt>
    2641. 

  2641. <dd class="field-even"><p>outputs, acc, net_time, gross_time
    2642. 

  2642. networks predictions, accuracy calculation, forward pass net time, function gross time</p>
    2643. 

  2643. </dd>
    2644. 

  2644. </dl>
    2645. 

  2645. </dd></dl>
    2646. 

  2646. 

  2647. <dl class="py property">
    2648. 

  2648. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.get_arch_params">
    2649. 

  2649. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_arch_params</span></span><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.get_arch_params" title="Permalink to this definition"></a></dt>
    2650. 

  2650. <dd></dd></dl>
    2651. 

  2651. 

  2652. <dl class="py property">
    2653. 

  2653. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.get_structure">
    2654. 

  2654. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_structure</span></span><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.get_structure" title="Permalink to this definition"></a></dt>
    2655. 

  2655. <dd></dd></dl>
    2656. 

  2656. 

  2657. <dl class="py property">
    2658. 

  2658. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.get_architecture">
    2659. 

  2659. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_architecture</span></span><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.get_architecture" title="Permalink to this definition"></a></dt>
    2660. 

  2660. <dd></dd></dl>
    2661. 

  2661. 

  2662. <dl class="py method">
    2663. 

  2663. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.set_experiment_name">
    2664. 

  2664. <span class="sig-name descname"><span class="pre">set_experiment_name</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">experiment_name</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_experiment_name"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.set_experiment_name" title="Permalink to this definition"></a></dt>
    2665. 

  2665. <dd></dd></dl>
    2666. 

  2666. 

  2667. <dl class="py property">
    2668. 

  2668. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.get_module">
    2669. 

  2669. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_module</span></span><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.get_module" title="Permalink to this definition"></a></dt>
    2670. 

  2670. <dd></dd></dl>
    2671. 

  2671. 

  2672. <dl class="py method">
    2673. 

  2673. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.set_module">
    2674. 

  2674. <span class="sig-name descname"><span class="pre">set_module</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/sg_model/sg_model.html#SgModel.set_module"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.set_module" title="Permalink to this definition"></a></dt>
    2675. 

  2675. <dd></dd></dl>
    2676. 

  2676. 

  2677. <dl class="py method">
    2678. 

  2678. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.test">
    2679. 

  2679. <span class="sig-name descname"><span class="pre">test</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">test_loader</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">torch.utils.data.dataloader.DataLoader</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">loss</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">torch.nn.modules.loss._Loss</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">silent_mode</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">test_metrics_list</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_logging_items_names</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_progress_verbose</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">test_phase_callbacks</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_ema_net</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> &#x2192; <span class="pre">tuple</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.test"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.test" title="Permalink to this definition"></a></dt>
    2680. 

  2680. <dd><p>Evaluates the model on given dataloader and metrics.</p>
    2681. 

  2681. <dl class="field-list simple">
    2682. 

  2682. <dt class="field-odd">Parameters</dt>
    2683. 

  2683. <dd class="field-odd"><ul class="simple">
    2684. 

  2684. <li><p><strong>test_loader</strong> – dataloader to perform test on.</p></li>
    2685. 

  2685. <li><p><strong>test_metrics_list</strong> – (list(torchmetrics.Metric)) metrics list for evaluation.</p></li>
    2686. 

  2686. <li><p><strong>silent_mode</strong> – (bool) controls verbosity</p></li>
    2687. 

  2687. <li><p><strong>metrics_progress_verbose</strong> – (bool) controls the verbosity of metrics progress (default=False). Slows down the program.</p></li>
    2688. 

  2688. </ul>
    2689. 

  2689. </dd>
    2690. 

  2690. </dl>
    2691. 

  2691. <dl class="simple">
    2692. 

  2692. <dt>:param use_ema_net (bool) whether to perform test on self.ema_model.ema (when self.ema_model.ema exists,</dt><dd><p>otherwise self.net will be tested) (default=True)</p>
    2693. 

  2693. </dd>
    2694. 

  2694. </dl>
    2695. 

  2695. <dl class="field-list simple">
    2696. 

  2696. <dt class="field-odd">Returns</dt>
    2697. 

  2697. <dd class="field-odd"><p>results tuple (tuple) containing the loss items and metric values.</p>
    2698. 

  2698. </dd>
    2699. 

  2699. </dl>
    2700. 

  2700. <dl class="simple">
    2701. 

  2701. <dt>All of the above args will override SgModel’s corresponding attribute when not equal to None. Then evaluation</dt><dd><p>is ran on self.test_loader with self.test_metrics.</p>
    2702. 

  2702. </dd>
    2703. 

  2703. </dl>
    2704. 

  2704. </dd></dl>
    2705. 

  2705. 

  2706. <dl class="py method">
    2707. 

  2707. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.evaluate">
    2708. 

  2708. <span class="sig-name descname"><span class="pre">evaluate</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data_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">metrics</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">torchmetrics.collections.MetricCollection</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">evaluation_type</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><a class="reference internal" href="super_gradients.common.data_types.enum.html#super_gradients.common.data_types.enum.EvaluationType" title="super_gradients.common.data_types.enum.evaluation_type.EvaluationType"><span class="pre">super_gradients.common.data_types.enum.evaluation_type.EvaluationType</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">epoch</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">silent_mode</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">metrics_progress_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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.evaluate"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.evaluate" title="Permalink to this definition"></a></dt>
    2709. 

  2709. <dd><p>Evaluates the model on given dataloader and metrics.</p>
    2710. 

  2710. <dl class="field-list simple">
    2711. 

  2711. <dt class="field-odd">Parameters</dt>
    2712. 

  2712. <dd class="field-odd"><ul class="simple">
    2713. 

  2713. <li><p><strong>data_loader</strong> – dataloader to perform evaluataion on</p></li>
    2714. 

  2714. <li><p><strong>metrics</strong> – (MetricCollection) metrics for evaluation</p></li>
    2715. 

  2715. <li><p><strong>evaluation_type</strong> – (EvaluationType) controls which phase callbacks will be used (for example, on batch end,
    2716. 

  2716. when evaluation_type=EvaluationType.VALIDATION the Phase.VALIDATION_BATCH_END callbacks will be triggered)</p></li>
    2717. 

  2717. <li><p><strong>epoch</strong> – (int) epoch idx</p></li>
    2718. 

  2718. <li><p><strong>silent_mode</strong> – (bool) controls verbosity</p></li>
    2719. 

  2719. <li><p><strong>metrics_progress_verbose</strong> – (bool) controls the verbosity of metrics progress (default=False).
    2720. 

  2720. Slows down the program significantly.</p></li>
    2721. 

  2721. </ul>
    2722. 

  2722. </dd>
    2723. 

  2723. <dt class="field-even">Returns</dt>
    2724. 

  2724. <dd class="field-even"><p>results tuple (tuple) containing the loss items and metric values.</p>
    2725. 

  2725. </dd>
    2726. 

  2726. </dl>
    2727. 

  2727. </dd></dl>
    2728. 

  2728. 

  2729. <dl class="py property">
    2730. 

  2730. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.get_net">
    2731. 

  2731. <em class="property"><span class="pre">property</span> </em><span class="sig-name descname"><span class="pre">get_net</span></span><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.get_net" title="Permalink to this definition"></a></dt>
    2732. 

  2732. <dd><p>Getter for network.
    2733. 

  2733. :return: torch.nn.Module, self.net</p>
    2734. 

  2734. </dd></dl>
    2735. 

  2735. 

  2736. <dl class="py method">
    2737. 

  2737. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.set_net">
    2738. 

  2738. <span class="sig-name descname"><span class="pre">set_net</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><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_net"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.set_net" title="Permalink to this definition"></a></dt>
    2739. 

  2739. <dd><p>Setter for network.</p>
    2740. 

  2740. <dl class="field-list simple">
    2741. 

  2741. <dt class="field-odd">Parameters</dt>
    2742. 

  2742. <dd class="field-odd"><p><strong>net</strong> – torch.nn.Module, value to set net</p>
    2743. 

  2743. </dd>
    2744. 

  2744. <dt class="field-even">Returns</dt>
    2745. 

  2745. <dd class="field-even"><p></p>
    2746. 

  2746. </dd>
    2747. 

  2747. </dl>
    2748. 

  2748. </dd></dl>
    2749. 

  2749. 

  2750. <dl class="py method">
    2751. 

  2751. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.set_ckpt_best_name">
    2752. 

  2752. <span class="sig-name descname"><span class="pre">set_ckpt_best_name</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">ckpt_best_name</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/super_gradients/training/sg_model/sg_model.html#SgModel.set_ckpt_best_name"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.set_ckpt_best_name" title="Permalink to this definition"></a></dt>
    2753. 

  2753. <dd><p>Setter for best checkpoint filename.</p>
    2754. 

  2754. <dl class="field-list simple">
    2755. 

  2755. <dt class="field-odd">Parameters</dt>
    2756. 

  2756. <dd class="field-odd"><p><strong>ckpt_best_name</strong> – str, value to set ckpt_best_name</p>
    2757. 

  2757. </dd>
    2758. 

  2758. </dl>
    2759. 

  2759. </dd></dl>
    2760. 

  2760. 

  2761. <dl class="py method">
    2762. 

  2762. <dt class="sig sig-object py" id="super_gradients.training.sg_model.SgModel.set_ema">
    2763. 

  2763. <span class="sig-name descname"><span class="pre">set_ema</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">val</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/sg_model/sg_model.html#SgModel.set_ema"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.SgModel.set_ema" title="Permalink to this definition"></a></dt>
    2764. 

  2764. <dd><p>Setter for self.ema</p>
    2765. 

  2765. <dl class="field-list simple">
    2766. 

  2766. <dt class="field-odd">Parameters</dt>
    2767. 

  2767. <dd class="field-odd"><p><strong>val</strong> – bool, value to set ema</p>
    2768. 

  2768. </dd>
    2769. 

  2769. </dl>
    2770. 

  2770. </dd></dl>
    2771. 

  2771. 

  2772. </dd></dl>
    2773. 

  2773. 

  2774. <dl class="py class">
    2775. 

  2775. <dt class="sig sig-object py" id="super_gradients.training.sg_model.MultiGPUMode">
    2776. 

  2776. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.sg_model.</span></span><span class="sig-name descname"><span class="pre">MultiGPUMode</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/common/data_types/enum/multi_gpu_mode.html#MultiGPUMode"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.MultiGPUMode" title="Permalink to this definition"></a></dt>
    2777. 

  2777. <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>
    2778. 

  2778. <dl class="py attribute">
    2779. 

  2779. <dt class="sig sig-object py">
    2780. 

  2780. <span class="sig-name descname"><span class="pre">OFF</span>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; <span class="pre">-</span> <span class="pre">Single</span> <span class="pre">GPU</span> <span class="pre">Mode</span> <span class="pre">/</span> <span class="pre">CPU</span> <span class="pre">Mode</span></span></dt>
    2781. 

  2781. <dd></dd></dl>
    2782. 

  2782. 

  2783. <dl class="py attribute">
    2784. 

  2784. <dt class="sig sig-object py">
    2785. 

  2785. <span class="sig-name descname"><span class="pre">DATA_PARALLEL</span>&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160;&#160; <span class="pre">-</span> <span class="pre">Multiple</span> <span class="pre">GPUs,</span> <span class="pre">Synchronous</span></span></dt>
    2786. 

  2786. <dd></dd></dl>
    2787. 

  2787. 

  2788. <dl class="py attribute">
    2789. 

  2789. <dt class="sig sig-object py">
    2790. 

  2790. <span class="sig-name descname"><span class="pre">DISTRIBUTED_DATA_PARALLEL</span> <span class="pre">-</span> <span class="pre">Multiple</span> <span class="pre">GPUs,</span> <span class="pre">Asynchronous</span></span></dt>
    2791. 

  2791. <dd></dd></dl>
    2792. 

  2792. 

  2793. <dl class="py attribute">
    2794. 

  2794. <dt class="sig sig-object py" id="super_gradients.training.sg_model.MultiGPUMode.OFF">
    2795. 

  2795. <span class="sig-name descname"><span class="pre">OFF</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'Off'</span></em><a class="headerlink" href="#super_gradients.training.sg_model.MultiGPUMode.OFF" title="Permalink to this definition"></a></dt>
    2796. 

  2796. <dd></dd></dl>
    2797. 

  2797. 

  2798. <dl class="py attribute">
    2799. 

  2799. <dt class="sig sig-object py" id="super_gradients.training.sg_model.MultiGPUMode.DATA_PARALLEL">
    2800. 

  2800. <span class="sig-name descname"><span class="pre">DATA_PARALLEL</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'DP'</span></em><a class="headerlink" href="#super_gradients.training.sg_model.MultiGPUMode.DATA_PARALLEL" title="Permalink to this definition"></a></dt>
    2801. 

  2801. <dd></dd></dl>
    2802. 

  2802. 

  2803. <dl class="py attribute">
    2804. 

  2804. <dt class="sig sig-object py" id="super_gradients.training.sg_model.MultiGPUMode.DISTRIBUTED_DATA_PARALLEL">
    2805. 

  2805. <span class="sig-name descname"><span class="pre">DISTRIBUTED_DATA_PARALLEL</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'DDP'</span></em><a class="headerlink" href="#super_gradients.training.sg_model.MultiGPUMode.DISTRIBUTED_DATA_PARALLEL" title="Permalink to this definition"></a></dt>
    2806. 

  2806. <dd></dd></dl>
    2807. 

  2807. 

  2808. <dl class="py attribute">
    2809. 

  2809. <dt class="sig sig-object py" id="super_gradients.training.sg_model.MultiGPUMode.AUTO">
    2810. 

  2810. <span class="sig-name descname"><span class="pre">AUTO</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'AUTO'</span></em><a class="headerlink" href="#super_gradients.training.sg_model.MultiGPUMode.AUTO" title="Permalink to this definition"></a></dt>
    2811. 

  2811. <dd></dd></dl>
    2812. 

  2812. 

  2813. </dd></dl>
    2814. 

  2814. 

  2815. <dl class="py class">
    2816. 

  2816. <dt class="sig sig-object py" id="super_gradients.training.sg_model.StrictLoad">
    2817. 

  2817. <em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">super_gradients.training.sg_model.</span></span><span class="sig-name descname"><span class="pre">StrictLoad</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/common/data_types/enum/strict_load.html#StrictLoad"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#super_gradients.training.sg_model.StrictLoad" title="Permalink to this definition"></a></dt>
    2818. 

  2818. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">enum.Enum</span></code></p>
    2819. 

  2819. <dl>
    2820. 

  2820. <dt>Wrapper for adding more functionality to torch’s strict_load parameter in load_state_dict().</dt><dd><dl>
    2821. 

  2821. <dt>Attributes:</dt><dd><p>OFF              - Native torch “strict_load = off” behaviour. See nn.Module.load_state_dict() documentation for more details.
    2822. 

  2822. ON               - Native torch “strict_load = on” behaviour. See nn.Module.load_state_dict() documentation for more details.
    2823. 

  2823. NO_KEY_MATCHING  - Allows the usage of SuperGradient’s adapt_checkpoint function, which loads a checkpoint by matching each</p>
    2824. 

  2824. <blockquote>
    2825. 

  2825. <div><p>layer’s shapes (and bypasses the strict matching of the names of each layer (ie: disregards the state_dict key matching)).</p>
    2826. 

  2826. </div></blockquote>
    2827. 

  2827. </dd>
    2828. 

  2828. </dl>
    2829. 

  2829. </dd>
    2830. 

  2830. </dl>
    2831. 

  2831. <dl class="py attribute">
    2832. 

  2832. <dt class="sig sig-object py" id="super_gradients.training.sg_model.StrictLoad.OFF">
    2833. 

  2833. <span class="sig-name descname"><span class="pre">OFF</span></span><em class="property"> <span class="pre">=</span> <span class="pre">False</span></em><a class="headerlink" href="#super_gradients.training.sg_model.StrictLoad.OFF" title="Permalink to this definition"></a></dt>
    2834. 

  2834. <dd></dd></dl>
    2835. 

  2835. 

  2836. <dl class="py attribute">
    2837. 

  2837. <dt class="sig sig-object py" id="super_gradients.training.sg_model.StrictLoad.ON">
    2838. 

  2838. <span class="sig-name descname"><span class="pre">ON</span></span><em class="property"> <span class="pre">=</span> <span class="pre">True</span></em><a class="headerlink" href="#super_gradients.training.sg_model.StrictLoad.ON" title="Permalink to this definition"></a></dt>
    2839. 

  2839. <dd></dd></dl>
    2840. 

  2840. 

  2841. <dl class="py attribute">
    2842. 

  2842. <dt class="sig sig-object py" id="super_gradients.training.sg_model.StrictLoad.NO_KEY_MATCHING">
    2843. 

  2843. <span class="sig-name descname"><span class="pre">NO_KEY_MATCHING</span></span><em class="property"> <span class="pre">=</span> <span class="pre">'no_key_matching'</span></em><a class="headerlink" href="#super_gradients.training.sg_model.StrictLoad.NO_KEY_MATCHING" title="Permalink to this definition"></a></dt>
    2844. 

  2844. <dd></dd></dl>
    2845. 

  2845. 

  2846. </dd></dl>
    2847. 

  2847. 

  2848. </section>
    2849. 

  2849. <section id="module-super_gradients.training.utils">
    2850. 

  2850. <span id="super-gradients-training-utils-module"></span><h2>super_gradients.training.utils module<a class="headerlink" href="#module-super_gradients.training.utils" title="Permalink to this headline"></a></h2>
    2851. 

  2851. <dl class="py class">
    2852. 

  2852. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer">
    2853. 

  2853. <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>
    2854. 

  2854. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    2855. 

  2855. <p>A class to measure time handling both GPU &amp; CPU processes
    2856. 

  2856. Returns time in milliseconds</p>
    2857. 

  2857. <dl class="py method">
    2858. 

  2858. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer.start">
    2859. 

  2859. <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>
    2860. 

  2860. <dd></dd></dl>
    2861. 

  2861. 

  2862. <dl class="py method">
    2863. 

  2863. <dt class="sig sig-object py" id="super_gradients.training.utils.Timer.stop">
    2864. 

  2864. <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>
    2865. 

  2865. <dd></dd></dl>
    2866. 

  2866. 

  2867. </dd></dl>
    2868. 

  2868. 

  2869. <dl class="py class">
    2870. 

  2870. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct">
    2871. 

  2871. <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>
    2872. 

  2872. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
    2873. 

  2873. <dl class="py method">
    2874. 

  2874. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.set_schema">
    2875. 

  2875. <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>
    2876. 

  2876. <dd></dd></dl>
    2877. 

  2877. 

  2878. <dl class="py method">
    2879. 

  2879. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.override">
    2880. 

  2880. <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>
    2881. 

  2881. <dd></dd></dl>
    2882. 

  2882. 

  2883. <dl class="py method">
    2884. 

  2884. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.to_dict">
    2885. 

  2885. <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>
    2886. 

  2886. <dd></dd></dl>
    2887. 

  2887. 

  2888. <dl class="py method">
    2889. 

  2889. <dt class="sig sig-object py" id="super_gradients.training.utils.HpmStruct.validate">
    2890. 

  2890. <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>
    2891. 

  2891. <dd><p>Validate the current dict values according to the provided schema
    2892. 

  2892. :raises</p>
    2893. 

  2893. <blockquote>
    2894. 

  2894. <div><p><cite>AttributeError</cite> if schema was not set
    2895. 

  2895. <cite>jsonschema.exceptions.ValidationError</cite> if the instance is invalid
    2896. 

  2896. <cite>jsonschema.exceptions.SchemaError</cite> if the schema itselfis invalid</p>
    2897. 

  2897. </div></blockquote>
    2898. 

  2898. </dd></dl>
    2899. 

  2899. 

  2900. </dd></dl>
    2901. 

  2901. 

  2902. <dl class="py class">
    2903. 

  2903. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel">
    2904. 

  2904. <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>
    2905. 

  2905. <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.nn.modules.module.Module</span></code></p>
    2906. 

  2906. <dl class="py method">
    2907. 

  2907. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel.forward">
    2908. 

  2908. <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>
    2909. 

  2909. <dd><p>Defines the computation performed at every call.</p>
    2910. 

  2910. <p>Should be overridden by all subclasses.</p>
    2911. 

  2911. <div class="admonition note">
    2912. 

  2912. <p class="admonition-title">Note</p>
    2913. 

  2913. <p>Although the recipe for forward pass needs to be defined within
    2914. 

  2914. this function, one should call the <code class="xref py py-class docutils literal notranslate"><span class="pre">Module</span></code> instance afterwards
    2915. 

  2915. instead of this since the former takes care of running the
    2916. 

  2916. registered hooks while the latter silently ignores them.</p>
    2917. 

  2917. </div>
    2918. 

  2918. </dd></dl>
    2919. 

  2919. 

  2920. <dl class="py attribute">
    2921. 

  2921. <dt class="sig sig-object py" id="super_gradients.training.utils.WrappedModel.training">
    2922. 

  2922. <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>
    2923. 

  2923. <dd></dd></dl>
    2924. 

  2924. 

  2925. </dd></dl>
    2926. 

  2926. 

  2927. <dl class="py function">
    2928. 

  2928. <dt class="sig sig-object py" id="super_gradients.training.utils.convert_to_tensor">
    2929. 

  2929. <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>
    2930. 

  2930. <dd><p>Converts numpy arrays and lists to Torch tensors before calculation losses
    2931. 

  2931. :param array: torch.tensor / Numpy array / List</p>
    2932. 

  2932. </dd></dl>
    2933. 

  2933. 

  2934. <dl class="py function">
    2935. 

  2935. <dt class="sig sig-object py" id="super_gradients.training.utils.get_param">
    2936. 

  2936. <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>
    2937. 

  2937. <dd><p>Retrieves a param from a parameter object/dict. If the parameter does not exist, will return default_val.
    2938. 

  2938. In case the default_val is of type dictionary, and a value is found in the params - the function
    2939. 

  2939. will return the default value dictionary with internal values overridden by the found value</p>
    2940. 

  2940. <p>i.e.
    2941. 

  2941. default_opt_params = {‘lr’:0.1, ‘momentum’:0.99, ‘alpha’:0.001}
    2942. 

  2942. training_params = {‘optimizer_params’: {‘lr’:0.0001}, ‘batch’: 32 …. }
    2943. 

  2943. get_param(training_params, name=’optimizer_params’, default_val=default_opt_params)
    2944. 

  2944. will return {‘lr’:0.0001, ‘momentum’:0.99, ‘alpha’:0.001}</p>
    2945. 

  2945. <dl class="field-list simple">
    2946. 

  2946. <dt class="field-odd">Parameters</dt>
    2947. 

  2947. <dd class="field-odd"><ul class="simple">
    2948. 

  2948. <li><p><strong>params</strong> – an object (typically HpmStruct) or a dict holding the params</p></li>
    2949. 

  2949. <li><p><strong>name</strong> – name of the searched parameter</p></li>
    2950. 

  2950. <li><p><strong>default_val</strong> – assumed to be the same type as the value searched in the params</p></li>
    2951. 

  2951. </ul>
    2952. 

  2952. </dd>
    2953. 

  2953. <dt class="field-even">Returns</dt>
    2954. 

  2954. <dd class="field-even"><p>the found value, or default if not found</p>
    2955. 

  2955. </dd>
    2956. 

  2956. </dl>
    2957. 

  2957. </dd></dl>
    2958. 

  2958. 

  2959. <dl class="py function">
    2960. 

  2960. <dt class="sig sig-object py" id="super_gradients.training.utils.tensor_container_to_device">
    2961. 

  2961. <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>
    2962. 

  2962. <dd><dl>
    2963. 

  2963. <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)
    2964. 

  2964. :param device:       device to send the tensors to
    2965. 

  2965. :param non_blocking: used for DistributedDataParallel
    2966. 

  2966. :returns        an object with the same structure (tensors, lists, tuples) with the device pointers (like</p>
    2967. 

  2967. <blockquote>
    2968. 

  2968. <div><p>the return value of Tensor.to(device)</p>
    2969. 

  2969. </div></blockquote>
    2970. 

  2970. </dd>
    2971. 

  2971. </dl>
    2972. 

  2972. </dd></dl>
    2973. 

  2973. 

  2974. <dl class="py function">
    2975. 

  2975. <dt class="sig sig-object py" id="super_gradients.training.utils.adapt_state_dict_to_fit_model_layer_names">
    2976. 

  2976. <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>, <em class="sig-param"><span class="n"><span class="pre">solver</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">callable</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/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>
    2977. 

  2977. <dd><p>Given a model state dict and source checkpoints, the method tries to correct the keys in the model_state_dict to fit
    2978. 

  2978. the ckpt in order to properly load the weights into the model. If unsuccessful - returns None</p>
    2979. 

  2979. <blockquote>
    2980. 

  2980. <div><dl class="field-list simple">
    2981. 

  2981. <dt class="field-odd">param model_state_dict</dt>
    2982. 

  2982. <dd class="field-odd"><p>the model state_dict</p>
    2983. 

  2983. </dd>
    2984. 

  2984. <dt class="field-even">param source_ckpt</dt>
    2985. 

  2985. <dd class="field-even"><p>checkpoint dict</p>
    2986. 

  2986. </dd>
    2987. 

  2987. </dl>
    2988. 

  2988. <p>:param exclude                  optional list for excluded layers
    2989. 

  2989. :param solver:                  callable with signature (ckpt_key, ckpt_val, model_key, model_val)</p>
    2990. 

  2990. <blockquote>
    2991. 

  2991. <div><p>that returns a desired weight for ckpt_val.</p>
    2992. 

  2992. </div></blockquote>
    2993. 

  2993. <dl class="field-list simple">
    2994. 

  2994. <dt class="field-odd">return</dt>
    2995. 

  2995. <dd class="field-odd"><p>renamed checkpoint dict (if possible)</p>
    2996. 

  2996. </dd>
    2997. 

  2997. </dl>
    2998. 

  2998. </div></blockquote>
    2999. 

  2999. </dd></dl>
    3000. 

  3000. 

  3001. <dl class="py function">
    3002. 

  3002. <dt class="sig sig-object py" id="super_gradients.training.utils.raise_informative_runtime_error">
    3003. 

  3003. <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>
    3004. 

  3004. <dd><p>Given a model state dict and source checkpoints, the method calls “adapt_state_dict_to_fit_model_layer_names”
    3005. 

  3005. and enhances the exception_msg if loading the checkpoint_dict via the conversion method is possible</p>
    3006. 

  3006. </dd></dl>
    3007. 

  3007. 

  3008. <dl class="py function">
    3009. 

  3009. <dt class="sig sig-object py" id="super_gradients.training.utils.random_seed">
    3010. 

  3010. <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>
    3011. 

  3011. <dd><p>Sets random seed of numpy, torch and random.</p>
    3012. 

  3012. <p>When using ddp a seed will be set for each process according to its local rank derived from the device number.
    3013. 

  3013. :param is_ddp: bool, will set different random seed for each process when using ddp.
    3014. 

  3014. :param device: ‘cuda’,’cpu’, ‘cuda:&lt;device_number&gt;’
    3015. 

  3015. :param seed: int, random seed to be set</p>
    3016. 

  3016. </dd></dl>
    3017. 

  3017. 

  3018. </section>
    3019. 

  3019. <section id="module-contents">
    3020. 

  3020. <h2>Module contents<a class="headerlink" href="#module-contents" title="Permalink to this headline"></a></h2>
    3021. 

  3021. </section>
    3022. 

  3022. </section>
    3023. 

  3023. 

  3024. 

  3025.            </div>
    3026. 

  3026.           </div>
    3027. 

  3027.           <footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
    3028. 

  3028.         <a href="super_gradients.common.html" class="btn btn-neutral float-left" title="Common package" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
    3029. 

  3029.         <a href="user_guide.html" class="btn btn-neutral float-right" title="What is SuperGradients?" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
    3030. 

  3030.     </div>
    3031. 

  3031. 

  3032.   <hr/>
    3033. 

  3033. 

  3034.   <div role="contentinfo">
    3035. 

  3035.     <p>&#169; Copyright 2021, SuperGradients team.</p>
    3036. 

  3036.   </div>
    3037. 

  3037. 

  3038.   Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
    3039. 

  3039.     <a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
    3040. 

  3040.     provided by <a href="https://readthedocs.org">Read the Docs</a>.
    3041. 

  3041.    
    3042. 

  3042. 

  3043. </footer>
    3044. 

  3044.         </div>
    3045. 

  3045.       </div>
    3046. 

  3046.     </section>
    3047. 

  3047.   </div>
    3048. 

  3048.   <script>
    3049. 

  3049.       jQuery(function () {
    3050. 

  3050.           SphinxRtdTheme.Navigation.enable(true);
    3051. 

  3051.       });
    3052. 

  3052.   </script> 
    3053. 

  3053. 

  3054. </body>
    3055. 

  3055. </html>
    3056.
Discard
Prev
Next
Tip!

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