Hugging Face's logo

wuff-mann
/
CNN-WaveAttention-v2

Model card Files
xet
 Community
wuff-mann commited on
Commit
4527038
·
verified ·
1 Parent(s): 73f1cf2

Upload logs/history.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. logs/history.json +340 -0
logs/history.json CHANGED
@@ -1358,5 +1358,345 @@
1358
  "diversity": 0.2354,
1359
  "grad_norm": 0.222,
1360
  "sec": 133.3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1361
  }
1362
  ]
 
1358
  "diversity": 0.2354,
1359
  "grad_norm": 0.222,
1360
  "sec": 133.3
1361
+ },
1362
+ {
1363
+ "epoch": 41,
1364
+ "beta": 0.0003,
1365
+ "lr": 2.43e-05,
1366
+ "train_loss": 0.043885,
1367
+ "train_recon": 0.035877,
1368
+ "train_kl": 26.6925,
1369
+ "val_loss": 0.042415,
1370
+ "val_recon": 0.034416,
1371
+ "val_kl": 26.6633,
1372
+ "active_dims": 18,
1373
+ "active_per_slot": [
1374
+ 3,
1375
+ 3,
1376
+ 3,
1377
+ 3,
1378
+ 3,
1379
+ 3
1380
+ ],
1381
+ "usage_max": 0.1701,
1382
+ "usage_min": 0.1633,
1383
+ "usage": [
1384
+ 0.1701,
1385
+ 0.169,
1386
+ 0.1666,
1387
+ 0.1633,
1388
+ 0.1668,
1389
+ 0.1643
1390
+ ],
1391
+ "active_slots": 6,
1392
+ "diversity": 0.2246,
1393
+ "grad_norm": 0.243,
1394
+ "sec": 133.8
1395
+ },
1396
+ {
1397
+ "epoch": 42,
1398
+ "beta": 0.0003,
1399
+ "lr": 1.95e-05,
1400
+ "train_loss": 0.043784,
1401
+ "train_recon": 0.035791,
1402
+ "train_kl": 26.6428,
1403
+ "val_loss": 0.042368,
1404
+ "val_recon": 0.034432,
1405
+ "val_kl": 26.453,
1406
+ "active_dims": 18,
1407
+ "active_per_slot": [
1408
+ 3,
1409
+ 3,
1410
+ 3,
1411
+ 3,
1412
+ 3,
1413
+ 3
1414
+ ],
1415
+ "usage_max": 0.1706,
1416
+ "usage_min": 0.1637,
1417
+ "usage": [
1418
+ 0.1706,
1419
+ 0.1642,
1420
+ 0.1637,
1421
+ 0.1677,
1422
+ 0.1664,
1423
+ 0.1674
1424
+ ],
1425
+ "active_slots": 6,
1426
+ "diversity": 0.2479,
1427
+ "grad_norm": 0.206,
1428
+ "sec": 132.2
1429
+ },
1430
+ {
1431
+ "epoch": 43,
1432
+ "beta": 0.0003,
1433
+ "lr": 1.52e-05,
1434
+ "train_loss": 0.043718,
1435
+ "train_recon": 0.035698,
1436
+ "train_kl": 26.7333,
1437
+ "val_loss": 0.042442,
1438
+ "val_recon": 0.034341,
1439
+ "val_kl": 27.0051,
1440
+ "active_dims": 18,
1441
+ "active_per_slot": [
1442
+ 3,
1443
+ 3,
1444
+ 3,
1445
+ 3,
1446
+ 3,
1447
+ 3
1448
+ ],
1449
+ "usage_max": 0.169,
1450
+ "usage_min": 0.1632,
1451
+ "usage": [
1452
+ 0.1637,
1453
+ 0.1685,
1454
+ 0.1632,
1455
+ 0.169,
1456
+ 0.1684,
1457
+ 0.1671
1458
+ ],
1459
+ "active_slots": 6,
1460
+ "diversity": 0.2401,
1461
+ "grad_norm": 0.216,
1462
+ "sec": 133.0
1463
+ },
1464
+ {
1465
+ "epoch": 44,
1466
+ "beta": 0.0003,
1467
+ "lr": 1.15e-05,
1468
+ "train_loss": 0.043645,
1469
+ "train_recon": 0.035621,
1470
+ "train_kl": 26.7465,
1471
+ "val_loss": 0.042475,
1472
+ "val_recon": 0.034462,
1473
+ "val_kl": 26.7104,
1474
+ "active_dims": 18,
1475
+ "active_per_slot": [
1476
+ 3,
1477
+ 3,
1478
+ 3,
1479
+ 3,
1480
+ 3,
1481
+ 3
1482
+ ],
1483
+ "usage_max": 0.1698,
1484
+ "usage_min": 0.1628,
1485
+ "usage": [
1486
+ 0.1634,
1487
+ 0.1673,
1488
+ 0.1628,
1489
+ 0.1681,
1490
+ 0.1698,
1491
+ 0.1686
1492
+ ],
1493
+ "active_slots": 6,
1494
+ "diversity": 0.2379,
1495
+ "grad_norm": 0.213,
1496
+ "sec": 133.1
1497
+ },
1498
+ {
1499
+ "epoch": 45,
1500
+ "beta": 0.0003,
1501
+ "lr": 8.3e-06,
1502
+ "train_loss": 0.043598,
1503
+ "train_recon": 0.035603,
1504
+ "train_kl": 26.6503,
1505
+ "val_loss": 0.042243,
1506
+ "val_recon": 0.034238,
1507
+ "val_kl": 26.6845,
1508
+ "active_dims": 18,
1509
+ "active_per_slot": [
1510
+ 3,
1511
+ 3,
1512
+ 3,
1513
+ 3,
1514
+ 3,
1515
+ 3
1516
+ ],
1517
+ "usage_max": 0.1682,
1518
+ "usage_min": 0.1643,
1519
+ "usage": [
1520
+ 0.1643,
1521
+ 0.1682,
1522
+ 0.1669,
1523
+ 0.1674,
1524
+ 0.1656,
1525
+ 0.1677
1526
+ ],
1527
+ "active_slots": 6,
1528
+ "diversity": 0.2375,
1529
+ "grad_norm": 0.278,
1530
+ "sec": 132.7
1531
+ },
1532
+ {
1533
+ "epoch": 46,
1534
+ "beta": 0.0003,
1535
+ "lr": 5.7e-06,
1536
+ "train_loss": 0.043559,
1537
+ "train_recon": 0.035543,
1538
+ "train_kl": 26.7212,
1539
+ "val_loss": 0.042285,
1540
+ "val_recon": 0.034253,
1541
+ "val_kl": 26.7713,
1542
+ "active_dims": 18,
1543
+ "active_per_slot": [
1544
+ 3,
1545
+ 3,
1546
+ 3,
1547
+ 3,
1548
+ 3,
1549
+ 3
1550
+ ],
1551
+ "usage_max": 0.1707,
1552
+ "usage_min": 0.1631,
1553
+ "usage": [
1554
+ 0.1703,
1555
+ 0.1665,
1556
+ 0.1654,
1557
+ 0.1707,
1558
+ 0.1631,
1559
+ 0.164
1560
+ ],
1561
+ "active_slots": 6,
1562
+ "diversity": 0.2418,
1563
+ "grad_norm": 0.206,
1564
+ "sec": 132.9
1565
+ },
1566
+ {
1567
+ "epoch": 47,
1568
+ "beta": 0.0003,
1569
+ "lr": 3.6e-06,
1570
+ "train_loss": 0.04352,
1571
+ "train_recon": 0.03552,
1572
+ "train_kl": 26.6669,
1573
+ "val_loss": 0.042189,
1574
+ "val_recon": 0.034201,
1575
+ "val_kl": 26.628,
1576
+ "active_dims": 18,
1577
+ "active_per_slot": [
1578
+ 3,
1579
+ 3,
1580
+ 3,
1581
+ 3,
1582
+ 3,
1583
+ 3
1584
+ ],
1585
+ "usage_max": 0.1708,
1586
+ "usage_min": 0.1624,
1587
+ "usage": [
1588
+ 0.1652,
1589
+ 0.1664,
1590
+ 0.1706,
1591
+ 0.1646,
1592
+ 0.1624,
1593
+ 0.1708
1594
+ ],
1595
+ "active_slots": 6,
1596
+ "diversity": 0.241,
1597
+ "grad_norm": 0.2,
1598
+ "sec": 132.8
1599
+ },
1600
+ {
1601
+ "epoch": 48,
1602
+ "beta": 0.0003,
1603
+ "lr": 2.2e-06,
1604
+ "train_loss": 0.043506,
1605
+ "train_recon": 0.035503,
1606
+ "train_kl": 26.6767,
1607
+ "val_loss": 0.042231,
1608
+ "val_recon": 0.034225,
1609
+ "val_kl": 26.6895,
1610
+ "active_dims": 18,
1611
+ "active_per_slot": [
1612
+ 3,
1613
+ 3,
1614
+ 3,
1615
+ 3,
1616
+ 3,
1617
+ 3
1618
+ ],
1619
+ "usage_max": 0.1696,
1620
+ "usage_min": 0.1648,
1621
+ "usage": [
1622
+ 0.1663,
1623
+ 0.1668,
1624
+ 0.1661,
1625
+ 0.1648,
1626
+ 0.1696,
1627
+ 0.1665
1628
+ ],
1629
+ "active_slots": 6,
1630
+ "diversity": 0.2376,
1631
+ "grad_norm": 0.235,
1632
+ "sec": 133.5
1633
+ },
1634
+ {
1635
+ "epoch": 49,
1636
+ "beta": 0.0003,
1637
+ "lr": 1.3e-06,
1638
+ "train_loss": 0.043493,
1639
+ "train_recon": 0.035485,
1640
+ "train_kl": 26.6949,
1641
+ "val_loss": 0.042185,
1642
+ "val_recon": 0.034197,
1643
+ "val_kl": 26.6262,
1644
+ "active_dims": 18,
1645
+ "active_per_slot": [
1646
+ 3,
1647
+ 3,
1648
+ 3,
1649
+ 3,
1650
+ 3,
1651
+ 3
1652
+ ],
1653
+ "usage_max": 0.1698,
1654
+ "usage_min": 0.1616,
1655
+ "usage": [
1656
+ 0.1616,
1657
+ 0.1678,
1658
+ 0.1683,
1659
+ 0.167,
1660
+ 0.1654,
1661
+ 0.1698
1662
+ ],
1663
+ "active_slots": 6,
1664
+ "diversity": 0.241,
1665
+ "grad_norm": 0.215,
1666
+ "sec": 133.1
1667
+ },
1668
+ {
1669
+ "epoch": 50,
1670
+ "beta": 0.0003,
1671
+ "lr": 1e-06,
1672
+ "train_loss": 0.043463,
1673
+ "train_recon": 0.035461,
1674
+ "train_kl": 26.6718,
1675
+ "val_loss": 0.042199,
1676
+ "val_recon": 0.034197,
1677
+ "val_kl": 26.6734,
1678
+ "active_dims": 18,
1679
+ "active_per_slot": [
1680
+ 3,
1681
+ 3,
1682
+ 3,
1683
+ 3,
1684
+ 3,
1685
+ 3
1686
+ ],
1687
+ "usage_max": 0.1688,
1688
+ "usage_min": 0.1648,
1689
+ "usage": [
1690
+ 0.1671,
1691
+ 0.1648,
1692
+ 0.1654,
1693
+ 0.1688,
1694
+ 0.1675,
1695
+ 0.1664
1696
+ ],
1697
+ "active_slots": 6,
1698
+ "diversity": 0.2411,
1699
+ "grad_norm": 0.212,
1700
+ "sec": 133.1
1701
  }
1702
  ]