Edit model card

wav2vec2-xls-r-300m-CV_Fleurs_AMMI_ALFFA-sw-100hrs-v1

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: 0.5053
  • Wer: 0.1682
  • Cer: 0.0545

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: 8
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • 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
2.1357 1.0 3941 0.5377 0.3748 0.1124
0.8699 2.0 7882 0.4594 0.3028 0.0896
0.7205 3.0 11823 0.4100 0.2756 0.0831
0.6396 4.0 15764 0.3702 0.2501 0.0760
0.5828 5.0 19705 0.3398 0.2445 0.0737
0.5376 6.0 23646 0.3592 0.2410 0.0778
0.5047 7.0 27587 0.3585 0.2302 0.0709
0.4787 8.0 31528 0.3120 0.2223 0.0678
0.453 9.0 35469 0.3507 0.2302 0.0771
0.429 10.0 39410 0.3253 0.2133 0.0669
0.4104 11.0 43351 0.3185 0.2130 0.0662
0.3901 12.0 47292 0.3281 0.2081 0.0647
0.3706 13.0 51233 0.3279 0.1988 0.0622
0.3573 14.0 55174 0.3187 0.2009 0.0625
0.341 15.0 59115 0.3253 0.2005 0.0618
0.3283 16.0 63056 0.3165 0.2009 0.0632
0.3173 17.0 66997 0.3499 0.1980 0.0623
0.3026 18.0 70938 0.3293 0.1968 0.0617
0.2913 19.0 74879 0.3176 0.1900 0.0607
0.2778 20.0 78820 0.3160 0.1915 0.0621
0.2678 21.0 82761 0.3431 0.1917 0.0607
0.2556 22.0 86702 0.3236 0.1855 0.0578
0.248 23.0 90643 0.3431 0.1883 0.0594
0.2377 24.0 94584 0.3600 0.1861 0.0585
0.2327 25.0 98525 0.3780 0.1879 0.0600
0.2247 26.0 102466 0.4006 0.1864 0.0596
0.2187 27.0 106407 0.4088 0.1883 0.0586
0.2088 28.0 110348 0.3501 0.1871 0.0591
0.2037 29.0 114289 0.3657 0.1803 0.0569
0.1977 30.0 118230 0.3654 0.1857 0.0596
0.1907 31.0 122171 0.4458 0.1844 0.0575
0.1878 32.0 126112 0.3657 0.1833 0.0591
0.1849 33.0 130053 0.4225 0.1800 0.0568
0.178 34.0 133994 0.4622 0.1794 0.0567
0.1728 35.0 137935 0.4237 0.1776 0.0564
0.1676 36.0 141876 0.4002 0.1776 0.0568
0.1651 37.0 145817 0.3979 0.1751 0.0557
0.159 38.0 149758 0.4132 0.1776 0.0569
0.155 39.0 153699 0.4201 0.1767 0.0561
0.1522 40.0 157640 0.4324 0.1724 0.0554
0.1499 41.0 161581 0.4637 0.1733 0.0553
0.1459 42.0 165522 0.4578 0.1756 0.0563
0.1416 43.0 169463 0.4288 0.1717 0.0553
0.1387 44.0 173404 0.4701 0.1739 0.0559
0.1353 45.0 177345 0.4323 0.1718 0.0546
0.13 46.0 181286 0.4630 0.1687 0.0546
0.1289 47.0 185227 0.4994 0.1693 0.0541
0.1244 48.0 189168 0.4498 0.1716 0.0545
0.1218 49.0 193109 0.4808 0.1715 0.0549
0.1193 50.0 197050 0.4820 0.1726 0.0559
0.1177 51.0 200991 0.4803 0.1687 0.0541
0.1133 52.0 204932 0.4955 0.1694 0.0545
0.1113 53.0 208873 0.4737 0.1677 0.0540
0.1085 54.0 212814 0.5311 0.1702 0.0546
0.1042 55.0 216755 0.4941 0.1682 0.0540
0.1065 56.0 220696 0.5053 0.1682 0.0545

Framework versions

  • Transformers 4.46.1
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.20.1
Downloads last month
15
Safetensors
Model size
316M 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_AMMI_ALFFA-sw-100hrs-v1

Finetuned
(439)
this model