wav2vec-xlsr-cv-grain-lg_both

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: 0.0871
  • Wer: 0.0289
  • Cer: 0.0079

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: 16
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • 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
1.9082 1.0 2703 0.2081 0.2630 0.0539
0.6172 2.0 5406 0.1523 0.1853 0.0386
0.5149 3.0 8109 0.1235 0.1432 0.0307
0.4589 4.0 10812 0.1207 0.1297 0.0285
0.4226 5.0 13515 0.1007 0.1117 0.0238
0.3932 6.0 16218 0.0988 0.0979 0.0225
0.3706 7.0 18921 0.0916 0.0924 0.0212
0.3481 8.0 21624 0.0882 0.0881 0.0210
0.3291 9.0 24327 0.0856 0.0803 0.0185
0.3131 10.0 27030 0.0806 0.0777 0.0182
0.2997 11.0 29733 0.0767 0.0733 0.0175
0.2851 12.0 32436 0.0749 0.0742 0.0178
0.2745 13.0 35139 0.0738 0.0636 0.0161
0.2628 14.0 37842 0.0818 0.0749 0.0177
0.2487 15.0 40545 0.0750 0.0699 0.0162
0.2389 16.0 43248 0.0682 0.0586 0.0145
0.2299 17.0 45951 0.0675 0.0590 0.0137
0.22 18.0 48654 0.0715 0.0562 0.0140
0.2102 19.0 51357 0.0757 0.0534 0.0134
0.202 20.0 54060 0.0785 0.0586 0.0149
0.195 21.0 56763 0.0731 0.0590 0.0136
0.186 22.0 59466 0.0750 0.0566 0.0136
0.1797 23.0 62169 0.0746 0.0523 0.0127
0.1713 24.0 64872 0.0739 0.0538 0.0132
0.1634 25.0 67575 0.0806 0.0514 0.0130
0.157 26.0 70278 0.0748 0.0532 0.0132
0.1521 27.0 72981 0.0774 0.0521 0.0133
0.1483 28.0 75684 0.0775 0.0501 0.0125
0.1424 29.0 78387 0.0772 0.0479 0.0122
0.1363 30.0 81090 0.0747 0.0453 0.0116
0.1322 31.0 83793 0.0801 0.0436 0.0116
0.1266 32.0 86496 0.0758 0.0473 0.0112
0.1234 33.0 89199 0.0691 0.0430 0.0107
0.1214 34.0 91902 0.0853 0.0458 0.0120
0.1178 35.0 94605 0.0805 0.0436 0.0107
0.1126 36.0 97308 0.0803 0.0436 0.0116
0.1119 37.0 100011 0.0744 0.0412 0.0105
0.1079 38.0 102714 0.0788 0.0421 0.0106
0.1039 39.0 105417 0.0802 0.0406 0.0105
0.1014 40.0 108120 0.0741 0.0367 0.0098
0.1 41.0 110823 0.0812 0.0401 0.0106
0.096 42.0 113526 0.0772 0.0410 0.0111
0.0937 43.0 116229 0.0782 0.0417 0.0106
0.0923 44.0 118932 0.0808 0.0404 0.0104
0.0894 45.0 121635 0.0725 0.0384 0.0097
0.0874 46.0 124338 0.0747 0.0351 0.0098
0.0856 47.0 127041 0.0761 0.0373 0.0100
0.0852 48.0 129744 0.0786 0.0393 0.0099
0.0821 49.0 132447 0.0766 0.0334 0.0092
0.0815 50.0 135150 0.0789 0.0365 0.0103
0.0798 51.0 137853 0.0813 0.0391 0.0101
0.0775 52.0 140556 0.0783 0.0341 0.0091
0.0755 53.0 143259 0.0793 0.0399 0.0105
0.0745 54.0 145962 0.0770 0.0408 0.0100
0.072 55.0 148665 0.0774 0.0349 0.0093
0.0708 56.0 151368 0.0811 0.0341 0.0091
0.0675 57.0 154071 0.0740 0.0321 0.0087
0.067 58.0 156774 0.0747 0.0321 0.0087
0.0657 59.0 159477 0.0721 0.0312 0.0085
0.0645 60.0 162180 0.0701 0.0341 0.0089
0.0631 61.0 164883 0.0788 0.0358 0.0090
0.0623 62.0 167586 0.0763 0.0312 0.0091
0.0614 63.0 170289 0.0777 0.0332 0.0087
0.0592 64.0 172992 0.0742 0.0319 0.0085
0.0576 65.0 175695 0.0755 0.0317 0.0085
0.0566 66.0 178398 0.0785 0.0347 0.0092
0.0565 67.0 181101 0.0794 0.0315 0.0086
0.0559 68.0 183804 0.0774 0.0317 0.0084
0.0534 69.0 186507 0.0814 0.0338 0.0088
0.0521 70.0 189210 0.0825 0.0330 0.0089
0.0514 71.0 191913 0.0781 0.0297 0.0081
0.0489 72.0 194616 0.0802 0.0293 0.0079
0.0496 73.0 197319 0.0799 0.0330 0.0086
0.0474 74.0 200022 0.0806 0.0299 0.0080
0.0479 75.0 202725 0.0789 0.0284 0.0080
0.0461 76.0 205428 0.0797 0.0308 0.0079
0.044 77.0 208131 0.0788 0.0284 0.0078
0.0444 78.0 210834 0.0830 0.0304 0.0083
0.0429 79.0 213537 0.0826 0.0312 0.0085
0.0423 80.0 216240 0.0845 0.0317 0.0087
0.041 81.0 218943 0.0862 0.0323 0.0085
0.0399 82.0 221646 0.0844 0.0297 0.0083
0.0402 83.0 224349 0.0884 0.0308 0.0084
0.0389 84.0 227052 0.0853 0.0276 0.0079
0.0372 85.0 229755 0.0839 0.0325 0.0082
0.0367 86.0 232458 0.0851 0.0282 0.0078
0.0358 87.0 235161 0.0836 0.0297 0.0081
0.0355 88.0 237864 0.0860 0.0295 0.0083
0.0347 89.0 240567 0.0848 0.0291 0.0081
0.0334 90.0 243270 0.0832 0.0280 0.0079
0.033 91.0 245973 0.0848 0.0282 0.0079
0.0329 92.0 248676 0.0852 0.0286 0.0082
0.0317 93.0 251379 0.0851 0.0291 0.0080
0.0314 94.0 254082 0.0873 0.0291 0.0080
0.0313 95.0 256785 0.0869 0.0284 0.0079
0.0305 96.0 259488 0.0853 0.0291 0.0080
0.03 97.0 262191 0.0862 0.0280 0.0077
0.0299 98.0 264894 0.0865 0.0282 0.0078
0.0286 99.0 267597 0.0871 0.0289 0.0079

Framework versions

  • Transformers 4.46.1
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.20.1
Downloads last month
143
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 sulaimank/wav2vec-xlsr-cv-grain-lg_both

Finetuned
(476)
this model