wav2vec2-xls-r-300m-CV-Fleurs-lg-10hrs-v6

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 1.3042
  • Wer: 0.6058
  • Cer: 0.1381

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0003
  • train_batch_size: 4
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
3.1664 1.0 1025 2.5342 1.0 0.8840
1.8493 2.0 2050 1.2129 0.9929 0.3862
1.2976 3.0 3075 0.9944 0.9482 0.2930
1.1249 4.0 4100 0.9405 0.9249 0.2753
0.9905 5.0 5125 0.8406 0.9072 0.2535
0.8961 6.0 6150 0.8121 0.8699 0.2368
0.8101 7.0 7175 0.7705 0.8607 0.2273
0.728 8.0 8200 0.7325 0.8330 0.2163
0.6653 9.0 9225 0.7483 0.8089 0.2060
0.6063 10.0 10250 0.7343 0.7981 0.2018
0.5503 11.0 11275 0.7499 0.7667 0.1887
0.5016 12.0 12300 0.7474 0.7734 0.1917
0.4574 13.0 13325 0.7665 0.7479 0.1858
0.4227 14.0 14350 0.8014 0.7605 0.1866
0.3898 15.0 15375 0.8140 0.7618 0.1817
0.3664 16.0 16400 0.7794 0.7234 0.1758
0.3389 17.0 17425 0.8175 0.7290 0.1762
0.3155 18.0 18450 0.8647 0.7284 0.1799
0.2991 19.0 19475 0.8268 0.7134 0.1731
0.2825 20.0 20500 0.9408 0.7312 0.1756
0.2665 21.0 21525 0.9131 0.7307 0.1715
0.253 22.0 22550 0.9645 0.7242 0.1747
0.2354 23.0 23575 0.9436 0.7125 0.1699
0.231 24.0 24600 0.9521 0.7239 0.1702
0.2178 25.0 25625 0.9751 0.7076 0.1694
0.2086 26.0 26650 0.9704 0.6945 0.1689
0.2002 27.0 27675 0.9937 0.7077 0.1682
0.1968 28.0 28700 0.9523 0.6959 0.1682
0.1889 29.0 29725 1.0351 0.6908 0.1653
0.182 30.0 30750 1.0054 0.6933 0.1644
0.1723 31.0 31775 1.0039 0.6930 0.1646
0.1695 32.0 32800 1.0005 0.6855 0.1632
0.1633 33.0 33825 1.0273 0.6897 0.1633
0.1571 34.0 34850 1.0361 0.6850 0.1615
0.1573 35.0 35875 1.0092 0.6767 0.1604
0.1511 36.0 36900 1.0353 0.6816 0.1622
0.1469 37.0 37925 1.0394 0.6716 0.1618
0.1495 38.0 38950 1.1006 0.6804 0.1621
0.1411 39.0 39975 1.1300 0.6742 0.1603
0.1391 40.0 41000 1.0378 0.6801 0.1591
0.138 41.0 42025 1.0655 0.6679 0.1581
0.1304 42.0 43050 1.1279 0.6777 0.1594
0.1308 43.0 44075 1.0743 0.6786 0.1572
0.128 44.0 45100 1.1424 0.6683 0.1569
0.1261 45.0 46125 1.0351 0.6787 0.1596
0.1242 46.0 47150 1.1587 0.6656 0.1556
0.1179 47.0 48175 1.1617 0.6538 0.1555
0.1164 48.0 49200 1.1593 0.6604 0.1567
0.1137 49.0 50225 1.1450 0.6622 0.1554
0.1102 50.0 51250 1.1221 0.6593 0.1555
0.1107 51.0 52275 1.1194 0.6618 0.1538
0.1088 52.0 53300 1.1452 0.6503 0.1527
0.1078 53.0 54325 1.1679 0.6529 0.1529
0.1054 54.0 55350 1.1926 0.6437 0.1503
0.1012 55.0 56375 1.1483 0.6568 0.1531
0.1047 56.0 57400 1.1756 0.6544 0.1528
0.0958 57.0 58425 1.2168 0.6531 0.1512
0.0966 58.0 59450 1.1973 0.6383 0.1493
0.0966 59.0 60475 1.1830 0.6493 0.1511
0.0948 60.0 61500 1.2027 0.6438 0.1509
0.0888 61.0 62525 1.1959 0.6413 0.1498
0.0909 62.0 63550 1.2046 0.6507 0.1512
0.0888 63.0 64575 1.2052 0.6347 0.1487
0.0869 64.0 65600 1.2118 0.6324 0.1482
0.0846 65.0 66625 1.1967 0.6365 0.1480
0.0833 66.0 67650 1.1957 0.6323 0.1460
0.0827 67.0 68675 1.1928 0.6370 0.1470
0.0824 68.0 69700 1.2578 0.6416 0.1472
0.08 69.0 70725 1.2427 0.6284 0.1447
0.0787 70.0 71750 1.2061 0.6295 0.1462
0.0777 71.0 72775 1.2185 0.6315 0.1454
0.0736 72.0 73800 1.2454 0.6237 0.1445
0.0746 73.0 74825 1.2629 0.6298 0.1464
0.0735 74.0 75850 1.2398 0.6218 0.1428
0.0724 75.0 76875 1.2727 0.6269 0.1440
0.0698 76.0 77900 1.2327 0.6259 0.1439
0.0677 77.0 78925 1.2338 0.6213 0.1442
0.0699 78.0 79950 1.2755 0.6226 0.1442
0.0656 79.0 80975 1.2734 0.6237 0.1431
0.0622 80.0 82000 1.2733 0.6211 0.1427
0.0648 81.0 83025 1.2345 0.6274 0.1421
0.0626 82.0 84050 1.2670 0.6273 0.1430
0.0632 83.0 85075 1.2634 0.6150 0.1422
0.0611 84.0 86100 1.3266 0.6192 0.1418
0.0608 85.0 87125 1.2889 0.6153 0.1414
0.0581 86.0 88150 1.2808 0.6146 0.1406
0.0586 87.0 89175 1.3120 0.6142 0.1406
0.0575 88.0 90200 1.2701 0.6171 0.1409
0.0577 89.0 91225 1.2916 0.6116 0.1400
0.0569 90.0 92250 1.3074 0.6132 0.1401
0.0552 91.0 93275 1.3030 0.6115 0.1388
0.0563 92.0 94300 1.2719 0.6082 0.1387
0.0516 93.0 95325 1.2853 0.6078 0.1380
0.0523 94.0 96350 1.2953 0.6096 0.1389
0.0489 95.0 97375 1.3099 0.6097 0.1387
0.0513 96.0 98400 1.3082 0.6095 0.1388
0.0522 97.0 99425 1.3076 0.6097 0.1384
0.0498 98.0 100450 1.3003 0.6073 0.1383
0.0506 99.0 101475 1.3012 0.6067 0.1382
0.0491 100.0 102500 1.3042 0.6058 0.1381

Framework versions

  • Transformers 4.46.1
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.20.1
Downloads last month
44
Safetensors
Model size
315M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for asr-africa/wav2vec2-xls-r-300m-CV-Fleurs-lg-10hrs-v6

Finetuned
(524)
this model