all_accents_combined_4x79
This model is a fine-tuned version of zainulhakim/240801-wav2vec2-ASR-Global-All-Clients on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1243
- Wer: 0.0568
- Cer: 0.0209
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: 4
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.6407 | 12.6582 | 1000 | 0.5311 | 0.5653 | 0.0874 |
| 0.3386 | 25.3165 | 2000 | 0.3585 | 0.3864 | 0.0491 |
| 0.2369 | 37.9747 | 3000 | 0.3723 | 0.3182 | 0.0584 |
| 0.1899 | 50.6329 | 4000 | 0.3953 | 0.3324 | 0.0509 |
| 0.1554 | 63.2911 | 5000 | 0.4053 | 0.3040 | 0.0550 |
| 0.1416 | 75.9494 | 6000 | 0.2952 | 0.2273 | 0.0425 |
| 0.1114 | 88.6076 | 7000 | 0.2022 | 0.2131 | 0.0368 |
| 0.0953 | 101.2658 | 8000 | 0.2619 | 0.2102 | 0.0371 |
| 0.0903 | 113.9241 | 9000 | 0.3288 | 0.2074 | 0.0375 |
| 0.0739 | 126.5823 | 10000 | 0.2549 | 0.1676 | 0.0321 |
| 0.0656 | 139.2405 | 11000 | 0.2227 | 0.1705 | 0.0293 |
| 0.0586 | 151.8987 | 12000 | 0.2873 | 0.1875 | 0.0349 |
| 0.0549 | 164.5570 | 13000 | 0.2169 | 0.1392 | 0.0243 |
| 0.0487 | 177.2152 | 14000 | 0.2097 | 0.1392 | 0.0267 |
| 0.0525 | 189.8734 | 15000 | 0.2497 | 0.1278 | 0.0337 |
| 0.0443 | 202.5316 | 16000 | 0.2112 | 0.1278 | 0.0291 |
| 0.0379 | 215.1899 | 17000 | 0.1633 | 0.1392 | 0.0237 |
| 0.038 | 227.8481 | 18000 | 0.2188 | 0.1420 | 0.0373 |
| 0.0408 | 240.5063 | 19000 | 0.3014 | 0.1335 | 0.0297 |
| 0.0287 | 253.1646 | 20000 | 0.2378 | 0.1222 | 0.0285 |
| 0.0294 | 265.8228 | 21000 | 0.2474 | 0.1278 | 0.0278 |
| 0.0276 | 278.4810 | 22000 | 0.2216 | 0.1136 | 0.0270 |
| 0.0323 | 291.1392 | 23000 | 0.1315 | 0.0881 | 0.0242 |
| 0.0239 | 303.7975 | 24000 | 0.1592 | 0.1080 | 0.0292 |
| 0.0187 | 316.4557 | 25000 | 0.1695 | 0.1335 | 0.0245 |
| 0.0169 | 329.1139 | 26000 | 0.2834 | 0.1136 | 0.0329 |
| 0.0163 | 341.7722 | 27000 | 0.1301 | 0.0852 | 0.0224 |
| 0.0151 | 354.4304 | 28000 | 0.2210 | 0.1080 | 0.0268 |
| 0.0102 | 367.0886 | 29000 | 0.1428 | 0.0852 | 0.0183 |
| 0.0098 | 379.7468 | 30000 | 0.1792 | 0.0653 | 0.0248 |
| 0.0104 | 392.4051 | 31000 | 0.2045 | 0.0682 | 0.0182 |
| 0.0101 | 405.0633 | 32000 | 0.2616 | 0.0881 | 0.0274 |
| 0.0069 | 417.7215 | 33000 | 0.1305 | 0.0739 | 0.0227 |
| 0.0063 | 430.3797 | 34000 | 0.1328 | 0.0795 | 0.0227 |
| 0.0046 | 443.0380 | 35000 | 0.1473 | 0.0682 | 0.0218 |
| 0.0039 | 455.6962 | 36000 | 0.1711 | 0.0710 | 0.0224 |
| 0.0028 | 468.3544 | 37000 | 0.1243 | 0.0568 | 0.0209 |
| 0.0022 | 481.0127 | 38000 | 0.1762 | 0.0625 | 0.0231 |
| 0.0025 | 493.6709 | 39000 | 0.1892 | 0.0625 | 0.0225 |
Framework versions
- Transformers 4.43.3
- Pytorch 2.3.1+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1