117:
1711:
275:. It provides a flexible N-dimensional array or Tensor, which supports basic routines for indexing, slicing, transposing, type-casting, resizing, sharing storage and cloning. This object is used by most other packages and thus forms the core object of the library. The Tensor also supports mathematical operations like
1214:
Many packages other than the above official packages are used with Torch. These are listed in the torch cheatsheet. These extra packages provide a wide range of utilities such as parallelism, asynchronous input/output, image processing, and so on. They can be installed with
1747:
1536:
947:
methods for computing the loss and backpropagating gradients, respectively. Criteria are helpful to train neural network on classical tasks. Common criteria are the
1905:
1427:
652:
Objects created with the torch factory can also be serialized, as long as they do not contain references to objects that cannot be serialized, such as Lua
634:
27:
1910:
1740:
1556:
1545:
1529:
1263:
1900:
1814:
1733:
288:
308:
1895:
1522:
1874:
1405:
260:
1822:
1644:
1203:
630:
96:
1609:
1195:
614:
70:
1756:
1244:
701:
642:
638:
232:
154:
130:
737:
646:
1240:
296:
244:
134:
1781:
1714:
1291:
626:
1251:. It has been used to build hardware implementations for data flows like those found in neural networks.
1768:
741:
312:
1374:
1483:
1837:
1664:
1566:
224:
217:
1690:
292:
169:
1296:
1854:
1841:
948:
228:
26:
1680:
1634:
709:
1332:
1584:
220:
186:
175:
145:
1438:
963:. What follows is an example of a Lua function that can be iteratively called to train an
1796:
1685:
1488:
705:
116:
1352:
1869:
1589:
685:
621:
by providing various convenience functions which are used throughout its packages. The
316:
1406:
KDnuggets
Interview with Yann LeCun, Deep Learning Expert, Director of Facebook AI Lab
1889:
956:
928:
618:
240:
179:
1579:
1574:
1202:
package provides much more options in this respect, like momentum and weight decay
736:
make up the basic component modules. This modular interface provides first-order
304:
191:
40:
1416:
1672:
1629:
1460:
1219:, the Lua package manager which is also included with the Torch distribution.
34:
1386:
1864:
1849:
1654:
1594:
1471:
664:
can be serialized if it is wrapped by a table (or metatable) that provides
653:
1449:
1254:
Facebook has released a set of extension modules as open source software.
1804:
1649:
1514:
1216:
158:
1786:
1764:
1675:
1624:
1290:"Torch: a modular machine learning software library". 30 October 2002.
1268:
256:
1695:
1619:
1614:
1391:
1363:
1317:
1236:
236:
138:
1827:
1725:
1604:
248:
150:
105:
1776:
1599:
323:
300:
252:
1729:
1518:
1472:
1248:
1232:
162:
724:
to create complex task-tailored graphs. Simpler modules like
1228:
688:. It is divided into modular objects that share a common
1333:"Torch7: A Matlab-like Environment for Machine Learning"
740:. What follows is an example use-case for building a
1836:
1813:
1795:
1763:
1663:
1565:
1312:
708:, respectively. Modules can be joined using module
202:
197:
185:
168:
144:
126:
95:
69:
47:
33:
1507:
1285:
1283:
322:The following exemplifies using torch via its
1741:
1530:
8:
839:-- some hyperbolic tangent transfer function
637:is called, torch initializes and sets a Lua
19:
1484:"Facebook Open-Sources a Trove of AI Tools"
1748:
1734:
1726:
1537:
1523:
1515:
1194:class for training a neural network using
115:
18:
1295:
1279:
1439:IDIAP Research Institute : Torch
1340:Neural Information Processing Systems
1243:. Torch has been extended for use on
255:. Torch development moved in 2017 to
7:
1264:Comparison of deep learning software
1906:Lua (programming language) software
935:, which has a similar interface to
623:torch.class(classname, parentclass)
931:are implemented as sub-classes of
738:automatic gradient differentiation
14:
287:, statistical distributions like
1710:
1709:
1387:"Cheatsheet · torch/torch7 Wiki"
25:
1450:Torch-android GitHub repository
625:function can be used to create
76:7.0 / February 27, 2017
1911:Software using the BSD license
1875:List of applications using Lua
1353:Torch GitHub repository ReadMe
1227:Torch is used by the Facebook
1:
684:package is used for building
271:The core package of Torch is
809:-- 10 input, 25 hidden units
313:matrix–matrix multiplication
309:matrix–vector multiplication
1610:Microsoft Cognitive Toolkit
1461:Torch-ios GitHub repository
1196:stochastic gradient descent
645:, which makes the table an
615:object-oriented programming
259:, a port of the library to
53:; 21 years ago
1927:
1428:Yann Lecun's Facebook Page
700:method that allow them to
692:interface. Modules have a
243:algorithms implemented in
1901:Free statistical software
1705:
1552:
1364:PyTorch GitHub repository
959:criterion implemented in
951:criterion implemented in
91:
65:
24:
1241:Idiap Research Institute
981:
746:
613:package also simplifies
328:
249:Idiap Research Institute
247:. It was created by the
967:Module on input Tensor
1896:Deep learning software
1546:Deep learning software
641:with the user-defined
16:Deep learning software
1698:Deep Learning Toolbox
1375:PyTorch Documentation
742:multilayer perceptron
78:; 7 years ago
225:scientific computing
1691:Wolfram Mathematica
21:
1492:. 16 January 2015.
1311:Collobert, Ronan.
1192:StochasticGradient
1104:zeroGradParameters
949:mean squared error
229:scripting language
43:, Johnny Mariéthoz
35:Original author(s)
1883:
1882:
1723:
1722:
1032:ClassNLLCriterion
961:ClassNLLCriterion
227:framework, and a
211:
210:
51:October 2002
39:Ronan Collobert,
1918:
1750:
1743:
1736:
1727:
1713:
1712:
1539:
1532:
1525:
1516:
1511:
1510:
1508:Official website
1494:
1493:
1480:
1474:
1469:
1463:
1458:
1452:
1447:
1441:
1436:
1430:
1425:
1419:
1414:
1408:
1403:
1397:
1396:
1383:
1377:
1372:
1366:
1361:
1355:
1350:
1344:
1343:
1337:
1329:
1323:
1322:
1308:
1302:
1301:
1299:
1287:
1231:Research Group,
1201:
1193:
1186:
1183:
1180:
1177:
1174:
1173:updateParameters
1171:
1168:
1165:
1162:
1159:
1156:
1153:
1150:
1147:
1144:
1141:
1138:
1135:
1132:
1129:
1126:
1123:
1120:
1117:
1114:
1111:
1108:
1105:
1102:
1099:
1096:
1093:
1090:
1087:
1084:
1081:
1078:
1075:
1072:
1069:
1066:
1063:
1060:
1057:
1054:
1051:
1048:
1045:
1042:
1039:
1036:
1033:
1030:
1027:
1024:
1021:
1018:
1015:
1012:
1009:
1006:
1003:
1000:
997:
994:
991:
988:
985:
978:
974:
971:, target Tensor
970:
966:
962:
954:
946:
942:
938:
934:
924:
921:
918:
915:
912:
909:
906:
903:
900:
897:
894:
891:
888:
885:
882:
879:
876:
873:
870:
867:
864:
861:
858:
855:
852:
849:
846:
843:
840:
837:
834:
831:
828:
825:
822:
819:
816:
813:
810:
807:
804:
801:
798:
795:
792:
789:
786:
783:
780:
777:
774:
771:
768:
765:
762:
759:
756:
753:
750:
735:
731:
727:
723:
719:
715:
699:
695:
691:
683:
671:
667:
627:object factories
624:
612:
605:
602:
599:
596:
593:
590:
587:
584:
581:
578:
575:
572:
569:
566:
563:
560:
557:
554:
551:
548:
545:
542:
539:
536:
533:
530:
527:
524:
521:
518:
515:
512:
509:
506:
503:
500:
497:
494:
491:
488:
485:
482:
479:
476:
473:
470:
467:
464:
461:
458:
455:
452:
449:
446:
443:
440:
437:
434:
433:0.34010116549482
431:
428:
425:
422:
419:
416:
413:
410:
407:
404:
401:
398:
395:
392:
389:
386:
383:
380:
377:
374:
371:
368:
365:
362:
359:
356:
353:
350:
347:
344:
341:
338:
335:
332:
303:operations like
286:
282:
278:
274:
221:machine learning
207:
204:
176:machine learning
146:Operating system
119:
114:
111:
109:
107:
86:
84:
79:
61:
59:
54:
29:
22:
1926:
1925:
1921:
1920:
1919:
1917:
1916:
1915:
1886:
1885:
1884:
1879:
1832:
1809:
1797:Package manager
1791:
1759:
1757:Lua programming
1754:
1724:
1719:
1701:
1686:Neural Designer
1659:
1561:
1548:
1543:
1506:
1505:
1502:
1497:
1482:
1481:
1477:
1470:
1466:
1459:
1455:
1448:
1444:
1437:
1433:
1426:
1422:
1415:
1411:
1404:
1400:
1385:
1384:
1380:
1373:
1369:
1362:
1358:
1351:
1347:
1335:
1331:
1330:
1326:
1310:
1309:
1305:
1289:
1288:
1281:
1277:
1260:
1225:
1212:
1199:
1198:, although the
1191:
1188:
1187:
1184:
1181:
1178:
1175:
1172:
1169:
1166:
1163:
1160:
1157:
1154:
1151:
1148:
1145:
1142:
1139:
1136:
1133:
1130:
1127:
1124:
1121:
1118:
1115:
1112:
1109:
1106:
1103:
1100:
1097:
1094:
1091:
1088:
1085:
1082:
1079:
1076:
1073:
1070:
1067:
1064:
1061:
1058:
1055:
1052:
1049:
1046:
1043:
1040:
1037:
1034:
1031:
1028:
1025:
1022:
1019:
1016:
1013:
1010:
1007:
1004:
1001:
998:
995:
992:
989:
986:
983:
976:
972:
968:
964:
960:
952:
944:
940:
936:
932:
926:
925:
922:
919:
916:
913:
910:
907:
904:
901:
898:
895:
892:
889:
886:
883:
880:
877:
874:
871:
868:
865:
862:
859:
856:
853:
850:
847:
844:
841:
838:
835:
832:
829:
826:
823:
820:
817:
814:
811:
808:
805:
802:
799:
796:
793:
790:
787:
784:
781:
778:
775:
772:
769:
766:
763:
760:
757:
754:
751:
748:
744:using Modules:
733:
729:
725:
721:
717:
713:
697:
693:
689:
686:neural networks
681:
678:
669:
665:
622:
610:
607:
606:
604:1.7844365427828
603:
600:
597:
594:
591:
588:
585:
582:
579:
576:
573:
570:
567:
564:
561:
558:
555:
552:
549:
546:
543:
540:
537:
534:
531:
528:
525:
522:
519:
516:
513:
510:
507:
504:
501:
498:
495:
492:
489:
486:
483:
480:
477:
474:
471:
468:
465:
462:
459:
456:
453:
450:
447:
444:
441:
438:
435:
432:
429:
426:
423:
420:
417:
414:
411:
408:
405:
402:
399:
396:
393:
390:
387:
384:
381:
378:
375:
372:
369:
366:
363:
360:
357:
354:
351:
348:
345:
342:
339:
336:
333:
330:
284:
280:
276:
272:
269:
201:
122:
104:
87:
82:
80:
77:
57:
55:
52:
48:Initial release
17:
12:
11:
5:
1924:
1922:
1914:
1913:
1908:
1903:
1898:
1888:
1887:
1881:
1880:
1878:
1877:
1872:
1867:
1862:
1857:
1852:
1846:
1844:
1834:
1833:
1831:
1830:
1825:
1819:
1817:
1811:
1810:
1808:
1807:
1801:
1799:
1793:
1792:
1790:
1789:
1784:
1779:
1773:
1771:
1761:
1760:
1755:
1753:
1752:
1745:
1738:
1730:
1721:
1720:
1718:
1717:
1706:
1703:
1702:
1700:
1699:
1693:
1688:
1683:
1678:
1669:
1667:
1661:
1660:
1658:
1657:
1652:
1647:
1642:
1637:
1632:
1627:
1622:
1617:
1612:
1607:
1602:
1597:
1592:
1590:Deeplearning4j
1587:
1582:
1577:
1571:
1569:
1563:
1562:
1560:
1559:
1553:
1550:
1549:
1544:
1542:
1541:
1534:
1527:
1519:
1513:
1512:
1501:
1500:External links
1498:
1496:
1495:
1475:
1464:
1453:
1442:
1431:
1420:
1409:
1398:
1378:
1367:
1356:
1345:
1324:
1303:
1278:
1276:
1273:
1272:
1271:
1266:
1259:
1256:
1224:
1221:
1211:
1210:Other packages
1208:
1204:regularization
982:
975:with a scalar
939:. It also has
929:Loss functions
747:
677:
674:
329:
317:matrix product
268:
265:
239:interfaces to
235:. It provides
209:
208:
199:
195:
194:
189:
183:
182:
172:
166:
165:
148:
142:
141:
128:
124:
123:
121:
120:
101:
99:
93:
92:
89:
88:
75:
73:
67:
66:
63:
62:
49:
45:
44:
37:
31:
30:
15:
13:
10:
9:
6:
4:
3:
2:
1923:
1912:
1909:
1907:
1904:
1902:
1899:
1897:
1894:
1893:
1891:
1876:
1873:
1871:
1868:
1866:
1863:
1861:
1858:
1856:
1853:
1851:
1848:
1847:
1845:
1843:
1839:
1835:
1829:
1826:
1824:
1821:
1820:
1818:
1816:
1812:
1806:
1803:
1802:
1800:
1798:
1794:
1788:
1785:
1783:
1780:
1778:
1775:
1774:
1772:
1770:
1769:distributions
1766:
1762:
1758:
1751:
1746:
1744:
1739:
1737:
1732:
1731:
1728:
1716:
1708:
1707:
1704:
1697:
1694:
1692:
1689:
1687:
1684:
1682:
1679:
1677:
1674:
1671:
1670:
1668:
1666:
1662:
1656:
1653:
1651:
1648:
1646:
1643:
1641:
1638:
1636:
1633:
1631:
1628:
1626:
1623:
1621:
1618:
1616:
1613:
1611:
1608:
1606:
1603:
1601:
1598:
1596:
1593:
1591:
1588:
1586:
1583:
1581:
1578:
1576:
1573:
1572:
1570:
1568:
1564:
1558:
1555:
1554:
1551:
1547:
1540:
1535:
1533:
1528:
1526:
1521:
1520:
1517:
1509:
1504:
1503:
1499:
1491:
1490:
1485:
1479:
1476:
1473:
1468:
1465:
1462:
1457:
1454:
1451:
1446:
1443:
1440:
1435:
1432:
1429:
1424:
1421:
1418:
1413:
1410:
1407:
1402:
1399:
1394:
1393:
1388:
1382:
1379:
1376:
1371:
1368:
1365:
1360:
1357:
1354:
1349:
1346:
1341:
1334:
1328:
1325:
1320:
1319:
1314:
1307:
1304:
1298:
1297:10.1.1.8.9850
1293:
1286:
1284:
1280:
1274:
1270:
1267:
1265:
1262:
1261:
1257:
1255:
1252:
1250:
1246:
1242:
1238:
1234:
1230:
1222:
1220:
1218:
1209:
1207:
1205:
1197:
980:
958:
957:cross-entropy
950:
930:
745:
743:
739:
711:
707:
706:backpropagate
703:
687:
675:
673:
663:
659:
655:
650:
648:
644:
640:
636:
632:
628:
620:
619:serialization
616:
327:
326:interpreter:
325:
320:
318:
314:
310:
306:
302:
298:
294:
290:
266:
264:
262:
258:
254:
250:
246:
242:
241:deep learning
238:
234:
230:
226:
222:
219:
215:
206:
200:
196:
193:
190:
188:
184:
181:
180:deep learning
177:
173:
171:
167:
164:
160:
156:
152:
149:
147:
143:
140:
136:
132:
129:
125:
118:
113:
103:
102:
100:
98:
94:
90:
74:
72:
71:Final release
68:
64:
50:
46:
42:
38:
36:
32:
28:
23:
1859:
1838:Applications
1639:
1580:Apache SINGA
1575:Apache MXNet
1487:
1478:
1467:
1456:
1445:
1434:
1423:
1412:
1401:
1390:
1381:
1370:
1359:
1348:
1339:
1327:
1316:
1306:
1253:
1226:
1223:Applications
1213:
1190:It also has
1189:
1179:learningRate
1011:learningRate
977:learningRate
953:MSECriterion
927:
679:
661:
657:
651:
633:). When the
608:
321:
270:
223:library, a
213:
212:
174:Library for
1665:Proprietary
1567:Open source
1417:Hacker News
881:-- 1 output
702:feedforward
660:. However,
635:constructor
305:dot product
297:multinomial
218:open-source
192:BSD License
41:Samy Bengio
1890:Categories
1842:frameworks
1681:IBM Watson
1630:TensorFlow
1557:Comparison
1275:References
987:gradUpdate
945:backward()
764:Sequential
714:Sequential
710:composites
698:backward()
656:, and Lua
654:coroutines
532:LongTensor
127:Written in
97:Repository
83:2017-02-27
1865:Tarantool
1850:OpenResty
1655:MindSpore
1595:DeepSpeed
1292:CiteSeerX
1119:criterion
1074:criterion
1020:criterion
941:forward()
933:Criterion
694:forward()
672:methods.
643:metatable
231:based on
1805:LuaRocks
1715:Category
1650:OpenVINO
1313:"Torch7"
1258:See also
1239:and the
1217:LuaRocks
1149:backward
1125:backward
984:function
955:and the
718:Parallel
662:userdata
658:userdata
159:Mac OS X
1855:Prosody
1787:Solar2D
1676:Core ML
1625:PyTorch
1342:. 2011.
1269:PyTorch
1245:Android
1080:forward
1053:forward
896:forward
712:, like
670:write()
631:classes
289:uniform
257:PyTorch
198:Website
187:License
155:Android
112:/torch7
81: (
58:2002-10
56: (
1823:Decoda
1696:MATLAB
1635:Theano
1620:OpenNN
1615:ML.NET
1392:GitHub
1318:GitHub
1294:
1237:Yandex
979::
937:Module
923:0.1815
863:Linear
791:Linear
726:Linear
722:Concat
690:Module
666:read()
647:object
583:0.8299
580:0.1708
577:1.6249
574:0.1411
571:0.2615
565:1.7844
559:0.3401
553:0.2381
502:0.8299
499:0.1708
496:1.6249
493:0.1411
490:0.2615
484:1.7844
478:0.3401
472:0.2381
445:narrow
421:0.8465
418:1.0525
415:2.2291
412:1.0434
406:0.8299
403:0.1708
400:1.6249
397:0.1411
394:0.2615
388:1.7844
382:0.3401
376:0.2381
299:, and
293:normal
261:Python
237:LuaJIT
216:is an
110:/torch
106:github
1860:Torch
1828:SciTE
1765:Ports
1673:Apple
1640:Torch
1605:Keras
1585:Caffe
1489:Wired
1336:(PDF)
1200:optim
1110:local
1065:local
1038:local
1017:local
908:randn
902:torch
639:table
611:torch
526:torch
514:index
346:randn
340:torch
273:torch
267:torch
214:Torch
203:torch
151:Linux
20:Torch
1815:IDEs
1782:LĂ–VE
1777:Plua
1645:ONNX
1600:Dlib
1247:and
1131:pred
1086:pred
1041:pred
943:and
884:>
842:>
833:Tanh
812:>
770:>
749:>
732:and
730:Tanh
720:and
704:and
696:and
680:The
668:and
617:and
609:The
586:>
505:>
436:>
424:>
364:>
331:>
324:REPL
315:and
301:BLAS
295:and
253:EPFL
178:and
170:Type
108:.com
1870:IUP
1249:iOS
1233:IBM
1185:end
1167:mlp
1143:mlp
1107:();
1098:mlp
1068:err
1047:mlp
993:mlp
965:mlp
890:mlp
851:add
845:mlp
836:())
821:add
815:mlp
779:add
773:mlp
752:mlp
734:Max
595:min
285:sum
281:min
277:max
251:at
233:Lua
205:.ch
163:iOS
139:C++
131:Lua
1892::
1840:,
1767:,
1486:.
1389:.
1338:.
1315:.
1282:^
1235:,
1229:AI
1206:.
1182:);
1164:);
1140:);
1095:);
1035:()
1026:nn
917:))
914:10
878:))
869:25
857:nn
827:nn
806:))
803:25
797:10
785:nn
767:()
758:nn
728:,
716:,
682:nn
676:nn
649:.
598:()
547:})
319:.
311:,
307:,
291:,
283:,
279:,
263:.
161:,
157:,
153:,
137:,
133:,
1749:e
1742:t
1735:v
1538:e
1531:t
1524:v
1395:.
1321:.
1300:.
1176:(
1170::
1161:t
1158:,
1155:x
1152:(
1146::
1137:y
1134:,
1128:(
1122::
1116:=
1113:t
1101::
1092:y
1089:,
1083:(
1077::
1071:=
1062:)
1059:x
1056:(
1050::
1044:=
1029:.
1023:=
1014:)
1008:,
1005:y
1002:,
999:x
996:,
990:(
973:y
969:x
920:-
911:(
905:.
899:(
893::
887:=
875:1
872:,
866:(
860:.
854:(
848::
830:.
824:(
818::
800:,
794:(
788:.
782:(
776::
761:.
755:=
629:(
601:-
592::
589:a
568:-
562:-
556:-
550:-
544:2
541:,
538:1
535:{
529:.
523:,
520:1
517:(
511::
508:a
487:-
481:-
475:-
469:-
466:)
463:2
460:,
457:1
454:,
451:1
448:(
442::
439:a
430:-
427:a
409:-
391:-
385:-
379:-
373:-
370:a
367:=
361:)
358:4
355:,
352:3
349:(
343:.
337:=
334:a
245:C
135:C
85:)
60:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.