Edit model card

wav2vec2-large-xlsr-53-german-cv8-dropout-30epoch

This model is a fine-tuned version of ./wav2vec2-large-xlsr-53-german-cv8-dropout-30epoch on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - DE dataset. It achieves the following results on the test set:

  • Wer: 10.29%
  • CER: 2.51%

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: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • num_epochs: 20.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.2081 1.0 6815 0.1784 0.1910
0.1686 2.0 13630 0.1621 0.1725
0.1515 3.0 20445 0.1569 0.1649
0.1426 4.0 27260 0.1466 0.1681
0.135 5.0 34075 0.1357 0.1410
0.1093 6.0 40890 0.1313 0.1436
0.1 7.0 47705 0.1242 0.1250
0.0999 8.0 54520 0.1191 0.1218
0.084 9.0 61335 0.1134 0.1164
0.0752 10.0 68150 0.1111 0.1117
0.0724 11.0 6815 0.1222 0.1206
0.0726 12.0 13630 0.1241 0.1247
0.0816 13.0 20445 0.1235 0.1174
0.0814 14.0 27260 0.1231 0.1238
0.063 15.0 34075 0.1171 0.1159
0.0793 16.0 40890 0.1158 0.1168
0.0686 17.0 47705 0.1187 0.1151
0.071 18.0 54520 0.1170 0.1182
0.0629 19.0 61335 0.1160 0.1085
0.0558 20.0 68150 0.1154 0.1093
0.0531 21.0 74965 0.1175 0.1044
0.0648 22.0 81780 0.1172 0.1056
0.0513 23.0 88595 0.1180 0.1048
0.0496 24.0 95410 0.1197 0.1025
0.0549 25.0 102225 0.1184 0.0991
0.0493 26.0 109040 0.1176 0.0977
0.0445 27.0 115855 0.1178 0.0989
0.0451 28.0 122670 0.1188 0.0992
0.045 29.0 129485 0.1182 0.0990
0.0452 30.0 136300 0.1190 0.0980

Framework versions

  • Transformers 4.19.0.dev0
  • Pytorch 1.11.0+cu113
  • Datasets 2.0.0
  • Tokenizers 0.11.6
Downloads last month
25
Hosted inference API
or or
This model can be loaded on the Inference API on-demand.

Dataset used to train oliverguhr/wav2vec2-large-xlsr-53-german

Evaluation results