<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-xlsr-ft-cy
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the CUSTOM_COMMON_VOICE.PY - CY dataset. It achieves the following results on the evaluation set:
- Loss: 0.1127
- Wer: 0.0622
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: 64
- eval_batch_size: 128
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 800
- num_epochs: 30.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
5.1868 | 0.68 | 400 | 0.7819 | 0.6673 |
0.4668 | 1.35 | 800 | 0.2890 | 0.3116 |
0.266 | 2.03 | 1200 | 0.1848 | 0.1822 |
0.1767 | 2.7 | 1600 | 0.1603 | 0.1521 |
0.1459 | 3.38 | 2000 | 0.1458 | 0.1401 |
0.1254 | 4.06 | 2400 | 0.1358 | 0.1280 |
0.1061 | 4.73 | 2800 | 0.1281 | 0.1195 |
0.0925 | 5.41 | 3200 | 0.1321 | 0.1224 |
0.0859 | 6.09 | 3600 | 0.1333 | 0.1181 |
0.0765 | 6.76 | 4000 | 0.1284 | 0.1148 |
0.0713 | 7.44 | 4400 | 0.1177 | 0.1046 |
0.0664 | 8.11 | 4800 | 0.1231 | 0.1015 |
0.0639 | 8.79 | 5200 | 0.1249 | 0.1007 |
0.0595 | 9.47 | 5600 | 0.1162 | 0.0922 |
0.0551 | 10.14 | 6000 | 0.1182 | 0.0917 |
0.0529 | 10.82 | 6400 | 0.1188 | 0.0903 |
0.0485 | 11.5 | 6800 | 0.1130 | 0.0886 |
0.0467 | 12.17 | 7200 | 0.1206 | 0.0953 |
0.0442 | 12.85 | 7600 | 0.1253 | 0.0964 |
0.0411 | 13.52 | 8000 | 0.1111 | 0.0889 |
0.0409 | 14.2 | 8400 | 0.1167 | 0.0854 |
0.0396 | 14.88 | 8800 | 0.1120 | 0.0864 |
0.0373 | 15.55 | 9200 | 0.1150 | 0.0807 |
0.0371 | 16.23 | 9600 | 0.1190 | 0.0801 |
0.0354 | 16.91 | 10000 | 0.1145 | 0.0795 |
0.0318 | 17.58 | 10400 | 0.1161 | 0.0759 |
0.0292 | 18.26 | 10800 | 0.1203 | 0.0765 |
0.028 | 18.93 | 11200 | 0.1187 | 0.0756 |
0.0273 | 19.61 | 11600 | 0.1254 | 0.0802 |
0.0275 | 20.29 | 12000 | 0.1171 | 0.0733 |
0.0264 | 20.96 | 12400 | 0.1116 | 0.0726 |
0.025 | 21.64 | 12800 | 0.1175 | 0.0726 |
0.0238 | 22.32 | 13200 | 0.1109 | 0.0697 |
0.0224 | 22.99 | 13600 | 0.1149 | 0.0711 |
0.0216 | 23.67 | 14000 | 0.1144 | 0.0693 |
0.0211 | 24.34 | 14400 | 0.1133 | 0.0685 |
0.0197 | 25.02 | 14800 | 0.1179 | 0.0671 |
0.0183 | 25.7 | 15200 | 0.1178 | 0.0684 |
0.0181 | 26.37 | 15600 | 0.1135 | 0.0655 |
0.0174 | 27.05 | 16000 | 0.1149 | 0.0647 |
0.0163 | 27.73 | 16400 | 0.1125 | 0.0628 |
0.0159 | 28.4 | 16800 | 0.1150 | 0.0630 |
0.0155 | 29.08 | 17200 | 0.1143 | 0.0621 |
0.0149 | 29.75 | 17600 | 0.1129 | 0.0621 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3