wav2vec2-xlsr-ln-50hr-v2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5435
  • Model Preparation Time: 0.0042
  • Wer: 0.1734
  • Cer: 0.0514

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: 0.0001
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Model Preparation Time Wer Cer
5.7126 0.9986 362 2.7144 0.0042 1.0 1.0
0.6867 2.0 725 0.4112 0.0042 0.3065 0.0896
0.232 2.9986 1087 0.3286 0.0042 0.2171 0.0670
0.1672 4.0 1450 0.2839 0.0042 0.1959 0.0573
0.1332 4.9986 1812 0.2715 0.0042 0.1884 0.0557
0.1104 6.0 2175 0.2864 0.0042 0.1778 0.0530
0.0955 6.9986 2537 0.2714 0.0042 0.1663 0.0498
0.0803 8.0 2900 0.2877 0.0042 0.1638 0.0495
0.0706 8.9986 3262 0.2848 0.0042 0.1689 0.0501
0.0628 10.0 3625 0.2903 0.0042 0.1651 0.0497
0.0582 10.9986 3987 0.2902 0.0042 0.1629 0.0490
0.0511 12.0 4350 0.3011 0.0042 0.1590 0.0482
0.046 12.9986 4712 0.3197 0.0042 0.1623 0.0483
0.0428 14.0 5075 0.3105 0.0042 0.1716 0.0499
0.0412 14.9986 5437 0.2977 0.0042 0.1554 0.0464
0.0373 16.0 5800 0.3222 0.0042 0.1516 0.0464
0.0337 16.9986 6162 0.3458 0.0042 0.1651 0.0517
0.0339 18.0 6525 0.3331 0.0042 0.1560 0.0467
0.0307 18.9986 6887 0.3184 0.0042 0.1519 0.0468
0.0299 20.0 7250 0.3591 0.0042 0.1592 0.0484
0.0275 20.9986 7612 0.3524 0.0042 0.1528 0.0474
0.0269 22.0 7975 0.3516 0.0042 0.1463 0.0450
0.0249 22.9986 8337 0.3370 0.0042 0.1533 0.0455
0.024 24.0 8700 0.3617 0.0042 0.1528 0.0456
0.0226 24.9986 9062 0.3661 0.0042 0.1520 0.0456
0.0218 26.0 9425 0.3470 0.0042 0.1523 0.0455
0.0196 26.9986 9787 0.3338 0.0042 0.1484 0.0449
0.0199 28.0 10150 0.3502 0.0042 0.1412 0.0444
0.0189 28.9986 10512 0.3429 0.0042 0.1446 0.0440
0.0187 30.0 10875 0.3500 0.0042 0.1452 0.0449
0.0181 30.9986 11237 0.3633 0.0042 0.1434 0.0438
0.0169 32.0 11600 0.3727 0.0042 0.1493 0.0451
0.0164 32.9986 11962 0.3554 0.0042 0.1567 0.0459
0.0168 34.0 12325 0.3715 0.0042 0.1435 0.044
0.0152 34.9986 12687 0.3579 0.0042 0.1409 0.0431
0.0143 36.0 13050 0.3873 0.0042 0.1429 0.0431
0.0148 36.9986 13412 0.3598 0.0042 0.1364 0.0421
0.0143 38.0 13775 0.3667 0.0042 0.1418 0.0440
0.0139 38.9986 14137 0.3864 0.0042 0.1330 0.0418
0.0139 40.0 14500 0.3431 0.0042 0.1368 0.0422
0.0125 40.9986 14862 0.3624 0.0042 0.1399 0.0430
0.0125 42.0 15225 0.3741 0.0042 0.1361 0.0432
0.0126 42.9986 15587 0.3678 0.0042 0.1402 0.0429
0.013 44.0 15950 0.3641 0.0042 0.1338 0.0412
0.012 44.9986 16312 0.3481 0.0042 0.1380 0.0422
0.0106 46.0 16675 0.3587 0.0042 0.1309 0.0404
0.0096 46.9986 17037 0.3667 0.0042 0.1305 0.0407
0.0101 48.0 17400 0.3897 0.0042 0.1319 0.0412
0.0109 48.9986 17762 0.3541 0.0042 0.1272 0.0409
0.0104 50.0 18125 0.3593 0.0042 0.1303 0.0397
0.0103 50.9986 18487 0.3572 0.0042 0.1300 0.0402
0.0097 52.0 18850 0.3740 0.0042 0.1305 0.0404
0.0092 52.9986 19212 0.3798 0.0042 0.1305 0.0403
0.0076 54.0 19575 0.3913 0.0042 0.1284 0.0400
0.0081 54.9986 19937 0.3684 0.0042 0.1335 0.0416
0.0084 56.0 20300 0.3895 0.0042 0.1277 0.0402
0.0079 56.9986 20662 0.3683 0.0042 0.1253 0.0394
0.0076 58.0 21025 0.3857 0.0042 0.1275 0.0401
0.0069 58.9986 21387 0.3922 0.0042 0.1243 0.0391
0.0069 60.0 21750 0.3913 0.0042 0.1278 0.0401
0.0065 60.9986 22112 0.4005 0.0042 0.1229 0.0396
0.0064 62.0 22475 0.3922 0.0042 0.1264 0.0396
0.0062 62.9986 22837 0.3959 0.0042 0.1261 0.0396
0.0059 64.0 23200 0.4036 0.0042 0.1233 0.0395
0.0063 64.9986 23562 0.4068 0.0042 0.1237 0.0392
0.0058 66.0 23925 0.4060 0.0042 0.1242 0.0391
0.0053 66.9986 24287 0.4147 0.0042 0.1220 0.0386
0.0057 68.0 24650 0.4002 0.0042 0.1217 0.0393
0.0051 68.9986 25012 0.4244 0.0042 0.1227 0.0393
0.005 70.0 25375 0.4146 0.0042 0.1221 0.0393
0.0051 70.9986 25737 0.4254 0.0042 0.1210 0.0391
0.0052 72.0 26100 0.4073 0.0042 0.1221 0.0391
0.005 72.9986 26462 0.4121 0.0042 0.1209 0.0388
0.0048 74.0 26825 0.4128 0.0042 0.1203 0.0389
0.0047 74.9986 27187 0.4114 0.0042 0.1209 0.0392
0.0044 76.0 27550 0.4068 0.0042 0.1218 0.0387
0.0042 76.9986 27912 0.4159 0.0042 0.1204 0.0384
0.0041 78.0 28275 0.4162 0.0042 0.1186 0.0383
0.004 78.9986 28637 0.4121 0.0042 0.1183 0.0379
0.0039 80.0 29000 0.4101 0.0042 0.1186 0.0381
0.0038 80.9986 29362 0.4109 0.0042 0.1174 0.0376
0.0037 82.0 29725 0.4115 0.0042 0.1183 0.0381
0.0035 82.9986 30087 0.4161 0.0042 0.1176 0.0379
0.0034 84.0 30450 0.4086 0.0042 0.1173 0.0379
0.0037 84.9986 30812 0.4133 0.0042 0.1180 0.0380
0.0033 86.0 31175 0.4163 0.0042 0.1176 0.0376
0.0036 86.9986 31537 0.4149 0.0042 0.1181 0.0378
0.0033 88.0 31900 0.4148 0.0042 0.1184 0.0378
0.0034 88.9986 32262 0.4142 0.0042 0.1169 0.0375
0.0032 90.0 32625 0.4198 0.0042 0.1168 0.0378
0.0036 90.9986 32987 0.4162 0.0042 0.1164 0.0376
0.0032 92.0 33350 0.4151 0.0042 0.1166 0.0377
0.0031 92.9986 33712 0.4175 0.0042 0.1167 0.0377
0.0033 94.0 34075 0.4170 0.0042 0.1167 0.0377
0.0031 94.9986 34437 0.4179 0.0042 0.1169 0.0377
0.003 96.0 34800 0.4179 0.0042 0.1170 0.0377
0.0032 96.9986 35162 0.4180 0.0042 0.1167 0.0377
0.0033 98.0 35525 0.4180 0.0042 0.1167 0.0377
0.0034 98.9986 35887 0.4179 0.0042 0.1169 0.0377
0.0032 99.8621 36200 0.4179 0.0042 0.1169 0.0377

Framework versions

  • Transformers 4.43.3
  • Pytorch 2.1.0+cu118
  • Datasets 2.20.0
  • Tokenizers 0.19.1
Downloads last month
160
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 KasuleTrevor/wav2vec2-xlsr-ln-50hr-v2

Finetuned
(524)
this model