Commit Graph

Select branches

Hide Pull Requests

conformer-ctc-readme

feature/lhotse-shar-example

gh-pages

master

streaming

#10

#100

#100

#1002

#1004

#1005

#1007

#1008

#1010

#1013

#1014

#1015

#1016

#1017

#1018

#1019

#1020

#1021

#1023

#1023

#1026

#1027

#1027

#1028

#1029

#1033

#1034

#1036

#1038

#1039

#104

#1043

#1044

#1046

#1047

#1048

#1049

#1050

#1051

#1052

#1053

#1053

#1055

#1057

#1058

#1059

#1060

#1061

#1061

#1066

#1067

#1070

#1072

#1074

#1075

#1076

#1077

#1078

#108

#1080

#1080

#1082

#1085

#1086

#109

#1093

#1095

#1096

#1097

#1099

#1099

#1101

#1102

#1104

#1105

#1106

#1106

#1107

#1108

#1109

#111

#1110

#1111

#1112

#1113

#1114

#1116

#1117

#1120

#1120

#1121

#1123

#1124

#1125

#1126

#1127

#1128

#1129

#113

#1130

#1131

#1132

#1133

#1135

#114

#1141

#1142

#1144

#1146

#1148

#115

#1150

#1152

#1153

#1157

#1158

#1159

#1160

#1161

#1162

#1164

#1165

#1166

#1167

#117

#1170

#1172

#1173

#1173

#1175

#1175

#1176

#1177

#1177

#1179

#118

#1180

#1181

#1183

#1185

#1186

#1187

#1188

#1189

#1190

#1190

#1191

#1193

#1194

#1197

#1198

#12

#120

#1200

#1202

#1204

#1207

#1208

#121

#1212

#1213

#1214

#1215

#1216

#1217

#122

#1220

#1222

#1222

#1226

#1226

#1229

#123

#1232

#1234

#1238

#1239

#124

#1240

#1241

#1242

#1243

#1244

#1248

#1249

#125

#1250

#1252

#1254

#1255

#1256

#1257

#1259

#1260

#1261

#1262

#1263

#1264

#1265

#1266

#1267

#1268

#1269

#127

#1270

#1272

#1273

#1275

#1277

#1278

#1279

#128

#1287

#129

#1290

#1291

#1291

#1292

#1293

#1296

#1297

#1299

#13

#1300

#1301

#1302

#1303

#1304

#1307

#1308

#131

#1310

#1310

#1313

#1314

#1316

#1317

#1318

#1319

#1321

#1322

#1324

#1325

#1326

#1329

#1330

#1332

#1333

#1334

#1336

#1337

#1338

#134

#1340

#1342

#1343

#1345

#1351

#1354

#1358

#1359

#1361

#1362

#1362

#1364

#1366

#1369

#1369

#137

#1372

#1374

#1376

#1376

#138

#138

#1380

#1381

#1385

#1386

#1389

#139

#1391

#1391

#1393

#1398

#14

#140

#1400

#1403

#1403

#1405

#1407

#1408

#141

#1410

#1411

#1412

#1413

#1415

#1416

#1421

#1422

#1424

#1425

#1427

#1428

#143

#1430

#1431

#1432

#1435

#1436

#1437

#1438

#1438

#1441

#1443

#1447

#1448

#1449

#145

#1450

#1455

#1460

#1464

#1466

#1467

#1468

#1469

#147

#1470

#1471

#1472

#1474

#1475

#1476

#148

#1482

#1483

#1484

#1485

#1487

#149

#1490

#1491

#1493

#1495

#1497

#1498

#1499

#15

#150

#150

#1500

#1501

#1502

#1503

#1504

#1509

#151

#1510

#1511

#1512

#1513

#1515

#1516

#152

#1520

#1520

#1521

#1522

#1524

#1526

#1527

#1528

#1529

#153

#1530

#1532

#1534

#1537

#1538

#1540

#1541

#1543

#1544

#1545

#1546

#1547

#1548

#155

#1550

#1551

#1554

#1555

#1556

#1557

#156

#1560

#1562

#1563

#1564

#1565

#1566

#1567

#1568

#157

#1571

#1571

#1573

#1575

#1577

#1578

#158

#158

#1582

#1582

#1583

#1583

#1584

#1585

#1586

#159

#159

#1590

#1593

#1593

#16

#160

#1601

#1602

#1603

#1604

#1605

#1607

#1609

#1611

#1613

#1617

#1619

#162

#1621

#1622

#1622

#1626

#1630

#1633

#1635

#164

#1643

#1645

#1645

#1646

#1647

#1648

#1649

#165

#1651

#1652

#1655

#1656

#1657

#166

#166

#1660

#1662

#1662

#1663

#1664

#1665

#1667

#1669

#1669

#167

#167

#1671

#1677

#1678

#1679

#168

#1681

#1682

#1683

#1684

#1684

#1686

#1687

#1689

#169

#1690

#1691

#1693

#1694

#17

#17

#170

#1700

#1704

#1706

#1707

#1707

#1708

#1708

#171

#1712

#1713

#1713

#1714

#1719

#172

#1721

#1722

#1727

#1730

#1732

#1732

#1734

#174

#1743

#1744

#1745

#1745

#1746

#1747

#1748

#1749

#1750

#1752

#1754

#1755

#1757

#1763

#1763

#1766

#1767

#1768

#1769

#1769

#177

#1770

#1772

#1773

#1774

#1775

#1776

#1778

#178

#178

#1781

#1782

#1785

#1786

#1787

#1788

#179

#1790

#1790

#1791

#1792

#1793

#1793

#1794

#1797

#18

#180

#1800

#1800

#1802

#1802

#1805

#1808

#1812

#1814

#1815

#1816

#1818

#1819

#182

#1820

#1821

#1821

#1825

#1827

#1828

#1829

#183

#183

#1830

#1835

#1837

#1838

#184

#184

#1840

#1841

#1845

#1845

#1846

#1849

#185

#1851

#1852

#1853

#1854

#1857

#1859

#186

#186

#1860

#1862

#1865

#1866

#1868

#187

#187

#1872

#1873

#1880

#1882

#1887

#1887

#1892

#1892

#1894

#1894

#19

#19

#190

#190

#1901

#1901

#1905

#191

#191

#1914

#1915

#1916

#1919

#192

#1926

#1929

#193

#1935

#1935

#1936

#194

#1940

#1941

#1942

#1942

#1944

#1944

#1947

#1949

#1950

#1952

#1954

#1955

#1959

#1959

#1964

#1965

#1966

#1967

#1969

#1973

#1974

#1975

#1975

#1976

#1977

#1979

#198

#1980

#1984

#1986

#1988

#199

#1990

#1991

#1992

#1992

#1995

#1997

#1997

#1999

#200

#200

#2014

#2015

#2015

#2019

#202

#2020

#2020

#2026

#2026

#2027

#204

#205

#205

#207

#208

#208

#21

#211

#213

#214

#215

#216

#216

#217

#218

#219

#22

#221

#222

#222

#223

#228

#229

#230

#231

#233

#234

#235

#236

#237

#239

#24

#241

#242

#242

#243

#244

#245

#246

#248

#25

#250

#251

#251

#252

#253

#254

#258

#259

#26

#260

#261

#262

#264

#265

#266

#267

#269

#27

#271

#272

#272

#274

#277

#278

#279

#28

#280

#281

#282

#283

#284

#285

#285

#287

#288

#289

#29

#291

#294

#295

#296

#298

#298

#299

#3

#30

#300

#301

#302

#303

#305

#307

#308

#309

#31

#310

#311

#312

#313

#314

#315

#316

#318

#321

#322

#323

#323

#325

#325

#326

#326

#327

#329

#330

#332

#333

#334

#336

#338

#339

#340

#343

#344

#345

#346

#346

#347

#348

#349

#350

#351

#352

#353

#354

#355

#356

#358

#359

#360

#361

#362

#363

#364

#365

#366

#367

#368

#369

#370

#371

#372

#373

#375

#376

#377

#378

#379

#38

#380

#382

#384

#386

#386

#387

#388

#389

#39

#390

#392

#395

#395

#396

#397

#398

#399

#4

#40

#400

#401

#402

#404

#407

#409

#41

#410

#411

#412

#413

#416

#417

#419

#419

#42

#420

#421

#425

#425

#427

#428

#429

#430

#433

#434

#435

#436

#437

#438

#439

#44

#440

#443

#444

#445

#447

#448

#448

#449

#45

#450

#451

#452

#453

#454

#456

#458

#458

#46

#460

#461

#462

#464

#465

#467

#468

#469

#470

#471

#472

#472

#475

#477

#477

#479

#481

#482

#483

#484

#485

#487

#488

#489

#490

#490

#492

#493

#493

#494

#495

#496

#497

#5

#50

#501

#504

#506

#507

#509

#509

#51

#512

#513

#514

#514

#516

#517

#518

#519

#52

#522

#523

#524

#525

#526

#527

#528

#529

#530

#530

#531

#532

#532

#533

#536

#537

#538

#539

#539

#54

#54

#540

#541

#542

#544

#545

#545

#546

#549

#55

#550

#551

#551

#552

#553

#554

#555

#558

#560

#560

#561

#562

#562

#563

#563

#564

#565

#565

#567

#568

#57

#571

#572

#573

#573

#574

#575

#58

#583

#584

#586

#588

#588

#591

#593

#595

#595

#597

#598

#6

#60

#601

#601

#604

#606

#609

#611

#612

#613

#614

#615

#617

#618

#619

#62

#621

#622

#623

#624

#624

#625

#627

#628

#629

#63

#63

#630

#631

#632

#635

#638

#639

#64

#640

#642

#645

#647

#648

#649

#65

#650

#653

#654

#656

#657

#659

#660

#662

#663

#663

#665

#668

#669

#670

#672

#675

#676

#678

#679

#679

#680

#681

#683

#686

#687

#688

#690

#691

#692

#693

#696

#698

#7

#700

#701

#704

#705

#706

#71

#717

#719

#72

#72

#720

#721

#721

#722

#725

#726

#727

#728

#729

#729

#73

#730

#731

#732

#732

#735

#737

#738

#742

#745

#745

#746

#75

#750

#751

#752

#753

#755

#758

#76

#762

#765

#768

#77

#773

#774

#778

#78

#782

#782

#783

#784

#787

#789

#79

#790

#791

#792

#795

#796

#797

#798

#799

#799

#8

#8

#80

#801

#804

#806

#808

#808

#81

#81

#812

#813

#815

#82

#820

#821

#822

#823

#824

#827

#828

#829

#829

#83

#830

#831

#832

#833

#835

#838

#84

#843

#844

#848

#849

#85

#852

#854

#856

#858

#86

#86

#861

#862

#863

#865

#868

#868

#869

#87

#870

#871

#874

#874

#875

#876

#879

#880

#881

#882

#883

#884

#888

#89

#890

#891

#892

#893

#894

#895

#897

#898

#9

#90

#900

#901

#902

#903

#904

#904

#905

#906

#907

#907

#908

#91

#91

#912

#913

#914

#915

#916

#919

#927

#933

#933

#934

#936

#937

#94

#941

#942

#943

#944

#945

#947

#949

#95

#95

#950

#950

#953

#954

#958

#958

#96

#961

#961

#962

#965

#967

#968

#969

#970

#971

#972

#974

#975

#976

#977

#98

#980

#981

#982

#983

#984

#985

#986

#988

#990

#992

#992

#993

#994

#995

#996

#997

v0.1

v1.0

v1.1

Mono Color

• 072776b2a1 Apply layer-skip dropout prob Daniel Povey 2022-10-29 20:11:39 +08:00
• 9a7979d7b8 Avoid falling off the loop for weird inputs Daniel Povey 2022-10-29 20:03:41 +08:00
• 05689f6354 Add skip connections as in normal U-net Daniel Povey 2022-10-29 19:47:10 +08:00
• bba454a349 Make decoder group size equal to 4. Daniel Povey 2022-10-29 17:08:19 +08:00
• ff03ec88a5 Tuning change to num encoder layers, inspired by relative param importance. Daniel Povey 2022-10-29 15:56:02 +08:00
• f995426324 Reduce min of bypass_scale from 0.5 to 0.3, and make it not applied in test mode. Daniel Povey 2022-10-29 15:40:47 +08:00
• 435d0dec71 Reduce dim of linear positional encoding in attention layers. Daniel Povey 2022-10-29 15:31:34 +08:00
• 96ea4cf1be Have 6 different encoder stacks, U-shaped network. Daniel Povey 2022-10-28 20:36:45 +08:00
• 7b57a34227 Have 4 encoder stacks Daniel Povey 2022-10-28 19:32:14 +08:00
• de9a6ebd6c Fix bug RE seq lengths Daniel Povey 2022-10-28 19:26:06 +08:00
• d7d5188bd9 Refactor how the downsampling is done so that it happens later, but the 1st encoder stack still operates after a subsampling of 2. Daniel Povey 2022-10-28 19:20:21 +08:00
• 0a89f51dc9 Have a 3rd encoder, at downsampling factor of 8. Daniel Povey 2022-10-28 17:39:29 +08:00
• ed1b4d5e5d Refactor zipformer for more flexibility so we can change number of encoder layers. Daniel Povey 2022-10-28 17:32:38 +08:00
• e592a920b4 Merge branch 'scaled_adam_exp198b' into scaled_adam_exp202 Daniel Povey 2022-10-28 13:13:55 +08:00
• a067fe8026 Fix clamping of epsilon Daniel Povey 2022-10-28 12:50:14 +08:00
• 7b8a0108ea Merge branch 'scaled_adam_exp188' into scaled_adam_exp198b Daniel Povey 2022-10-28 12:49:36 +08:00
• b9f6ba1aa2 Remove some unused variables. Daniel Povey 2022-10-28 12:01:45 +08:00
• c8abba75a9 Update decode.py by copying from pruned_transducer_stateless5 and changing directory name Daniel Povey 2022-10-28 11:19:45 +08:00
• 6709bf1e63
  Update train.py (#635) Nagendra Goel 2022-10-27 22:23:32 -04:00
• e08a44dd7f
  Merge pull request #1 from ngoel17/pruned-transducer3-train-bugfix Nagendra Goel 2022-10-27 22:10:13 -04:00
• 1ae0ee3104
  Update train.py Nagendra Goel 2022-10-27 22:08:02 -04:00
• 5dfa141ca5 Rename Conformer to Zipformer Daniel Povey 2022-10-27 22:43:46 +08:00
• 3f05e47447 Rename conformer.py to zipformer.py Daniel Povey 2022-10-27 22:41:48 +08:00
• be5c687fbd Merging upstream/master Daniel Povey 2022-10-27 21:04:48 +08:00
• f8c531cd23 Increase bypass_scale min from 0.4 to 0.5 Daniel Povey 2022-10-27 14:59:05 +08:00
• 2c400115e4 Increase bypass_scale from 0.2 to 0.4. Daniel Povey 2022-10-27 14:30:46 +08:00
• a7fc6ae38c Increase floor on bypass_scale from 0.1 to 0.2. Daniel Povey 2022-10-27 14:09:34 +08:00
• 938510ac9f Fix clamping of bypass scale; remove a couple unused variables. Daniel Povey 2022-10-27 14:05:53 +08:00
• bf37c7ca85 Regularize how we apply the min and max to the eps of BasicNorm Daniel Povey 2022-10-26 12:51:20 +08:00
• a0507a83a5 Change scalar_max in optim.py from 2.0 to 5.0 Daniel Povey 2022-10-25 22:45:08 +08:00
• 78f3cba58c Add logging about memory used. Daniel Povey 2022-10-25 19:19:33 +08:00
• 6a6df19bde Hopefully make penalize_abs_values_gt more memory efficient. Daniel Povey 2022-10-25 18:34:04 +08:00
• dbfbd8016b Cast to float16 in DoubleSwish forward Daniel Povey 2022-10-25 13:16:00 +08:00
• 3159b09e8f Make 20 the limit for warmup_count Daniel Povey 2022-10-25 12:58:27 +08:00
• 6ebff23cb9 Reduce cutoff from 100 to 5 for estimating OOM with warmup Daniel Povey 2022-10-25 12:53:12 +08:00
• 9da5526659 Changes to more accurately estimate OOM conditions Daniel Povey 2022-10-25 12:49:18 +08:00
• 1e8984174b Change to warmup schedule. Daniel Povey 2022-10-25 12:27:00 +08:00
• 36cb279318 More memory efficient backprop for DoubleSwish. Daniel Povey 2022-10-25 12:21:22 +08:00
• 499ac24ecb
  Install kaldifst for GitHub actions (#632) Fangjun Kuang 2022-10-24 15:07:29 +08:00
• bc0f3f260e Fix CI errors Fangjun Kuang 2022-10-24 11:30:53 +08:00
• eee48dd7ac Install kaldifst for GitHub actions Fangjun Kuang 2022-10-24 10:44:11 +08:00
• 95aaa4a8d2 Store only half precision output for softmax. Daniel Povey 2022-10-23 21:24:46 +08:00
• d3876e32c4 Make it use float16 if in amp but use clamp to avoid wrapping error Daniel Povey 2022-10-23 21:13:23 +08:00
• 85657946bb Try a more exact way to round to uint8 that should prevent ever wrapping around to zero Daniel Povey 2022-10-23 20:56:26 +08:00
• d6aa386552 Fix randn to rand Daniel Povey 2022-10-23 17:19:19 +08:00
• e586cc319c Change the discretization of the sigmoid to be expectation preserving. Daniel Povey 2022-10-23 17:11:35 +08:00
• 09cbc9fdab Save some memory in the autograd of DoubleSwish. Daniel Povey 2022-10-23 16:59:43 +08:00
• 40588d3d8a Revert 179->180 change, i.e. change max_abs for deriv_balancer2 back from 50.0 20.0 Daniel Povey 2022-10-23 16:18:58 +08:00
• c7ac266f9a
  Merge ebcc8e4e1de7f8e97ce1587264166dc865459371 into 348494888d08d5ddba2baadddcfe7df576d4bed1 kobenaxie 2022-10-23 15:55:35 +09:00
• 5b9d166cb9 --base-lr0.075->0.5; --lr-epochs 3->3.5 Daniel Povey 2022-10-23 13:56:25 +08:00
• 0406d0b059 Increase max_abs in ActivationBalancer of conv module from 20 to 50 Daniel Povey 2022-10-23 13:51:51 +08:00
• 9e86d1f44f reduce initial scale in GradScaler Daniel Povey 2022-10-23 00:14:38 +08:00
• b7083e7aff Increase default max_factor for ActivationBalancer from 0.02 to 0.04; decrease max_abs in ConvolutionModule.deriv_balancer2 from 100.0 to 20.0 Daniel Povey 2022-10-23 00:09:21 +08:00
• ad2d3c2b36 Dont print out full non-finite tensor Daniel Povey 2022-10-22 23:03:19 +08:00
• e0c1dc66da Increase probs of activation balancer and make it decay slower. Daniel Povey 2022-10-22 22:18:38 +08:00
• 2964628ae1 don't do penalize_values_gt on simple_lm_proj and simple_am_proj; reduce --base-lr from 0.1 to 0.075 Daniel Povey 2022-10-22 21:12:58 +08:00
• 269b70122e Add hooks.py, had negleted to git add it. Daniel Povey 2022-10-22 20:58:52 +08:00
• 13ffd8e823 Trying to reduce grad_scale of Whiten() from 0.02 to 0.01. Daniel Povey 2022-10-22 20:30:05 +08:00
• 466176eeff Use penalize_abs_values_gt, not ActivationBalancer. Daniel Povey 2022-10-22 20:15:21 +08:00
• 7a55cac346 Increase max_factor in final lm_balancer and am_balancer Daniel Povey 2022-10-22 20:02:54 +08:00
• 8b3bba9b54 Reduce max_abs on am_balancer Daniel Povey 2022-10-22 19:52:11 +08:00
• 1908123af9 Adding activation balancers after simple_am_prob and simple_lm_prob Daniel Povey 2022-10-22 19:37:35 +08:00
• 11886dc4f6 Change base lr to 0.1, also rename from initial lr in train.py Daniel Povey 2022-10-22 18:22:26 +08:00
• 146626bb85 Renaming in optim.py; remove step() from scan_pessimistic_batches_for_oom in train.py Daniel Povey 2022-10-22 17:44:21 +08:00
• 525e87a82d Add inf check hooks Daniel Povey 2022-10-22 17:16:29 +08:00
• e8066b5825 Merge branch 'scaled_adam_exp172' into scaled_adam_exp174 Daniel Povey 2022-10-22 15:44:04 +08:00
• 9919fb3e1b Increase grad_scale to Whiten module Daniel Povey 2022-10-22 15:32:50 +08:00
• af0fc31c78 Introduce warmup schedule in optimizer Daniel Povey 2022-10-22 15:15:43 +08:00
• 069125686e Fixes to logging statements. Daniel Povey 2022-10-22 15:08:07 +08:00
• 1d4382555c Increase initial-lr from 0.06 to 0.075 and decrease lr-epochs from 3.5 to 3. Daniel Povey 2022-10-22 15:04:08 +08:00
• 74d775014d Increase initial-lr from 0.05 to 0.06. Daniel Povey 2022-10-22 15:02:07 +08:00
• aa5f34af64 Cosmetic change Daniel Povey 2022-10-22 15:00:15 +08:00
• 1ec9fe5c98 Make warmup period decrease scale on simple loss, leaving pruned loss scale constant. Daniel Povey 2022-10-22 14:48:53 +08:00
• efde3757c7 Reset optimizer state when we change loss function definition. Daniel Povey 2022-10-22 14:30:18 +08:00
• 84580ec022 Configuration changes: scores limit 5->10, min_prob 0.05->0.1, cur_grad_scale more aggressive increase Daniel Povey 2022-10-22 14:09:53 +08:00
• 9672dffac2 Merge branch 'scaled_adam_exp168' into scaled_adam_exp169 Daniel Povey 2022-10-22 14:05:07 +08:00
• 8d1021d131 Remove comparison diagnostics, which were not that useful. Daniel Povey 2022-10-22 13:57:00 +08:00
• 2e93e5d3b7 Add logging Daniel Povey 2022-10-22 13:52:51 +08:00
• fd3f21f84d Changes to grad scale logging; increase grad scale more frequently if less than one. Daniel Povey 2022-10-22 13:36:26 +08:00
• 348494888d
  Add kaldifst to requirements.txt (#631) Fangjun Kuang 2022-10-22 13:14:44 +08:00
• 83cfc78c0e Add kaldifst to requirements.txt Fangjun Kuang 2022-10-22 13:13:23 +08:00
• 1d2fe8e3c2 Add more diagnostics to debug gradient scale problems Daniel Povey 2022-10-22 12:49:29 +08:00
• 3298e18732 Increase limit on logit for SimpleCombiner to 25.0 Daniel Povey 2022-10-21 22:06:05 +08:00
• e5fe3de17e Also apply limit on logit in SimpleCombiner Daniel Povey 2022-10-21 21:31:52 +08:00
• bdbd2cfce6 Penalize too large weights in softmax of AttentionDownsample() Daniel Povey 2022-10-21 20:12:36 +08:00
• 9b671e1c21
  Add Shallow fusion in modified_beam_search (#630) ezerhouni 2022-10-21 10:44:56 +02:00
• f5f3cf2869 Fix code according to review Erwan 2022-10-21 09:45:41 +02:00
• 476fb9e9f3 Reduce min_prob of ActivationBalancer from 0.1 to 0.05. Daniel Povey 2022-10-21 15:42:04 +08:00
• 13411dbee0 Modified beam search with ngram rescoring Erwan 2022-10-20 21:53:41 +02:00
• 9f68b5717c Reduce the limit on attention weights from 50 to 25. Daniel Povey 2022-10-21 12:13:23 +08:00
• 347031d12d revert utils pkufool 2022-10-21 11:50:31 +08:00
• 9c5488ee03 support streaming for aishell stateless2; add delay penalty to aishell pkufool 2022-10-21 11:39:28 +08:00
• 96bb44eb04 Use shallow fusion for modified-beam-search Fangjun Kuang 2022-10-18 23:42:11 +08:00
• b7782bbe0c test batch size == 1 without shallow fusion Fangjun Kuang 2022-10-17 17:29:58 +08:00
• 2802ef7700 Add utility for shallow fusion Fangjun Kuang 2022-10-10 09:20:28 +08:00
• c5cb52fed1 Remove the use of random_clamp in conformer.py. Daniel Povey 2022-10-20 19:54:38 +08:00
• dccff6b893 Remove use of RandomGrad Daniel Povey 2022-10-20 19:35:11 +08:00
• 1018a77410 Use normal implementation of softmax. Daniel Povey 2022-10-20 19:34:10 +08:00
• 6e6209419c Merge branch 'scaled_adam_exp150' into scaled_adam_exp155 Daniel Povey 2022-10-20 15:04:27 +08:00
• 4565d43d5c Add hard limit of attention weights to +- 50 Daniel Povey 2022-10-20 14:27:55 +08:00