<!-- 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. -->
xlsr-syntesized-turkish-8-hour
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1542
- Wer: 0.0998
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: 0.0005
- train_batch_size: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
4.6973 | 0.26 | 100 | 3.0648 | 1.0 |
2.1789 | 0.52 | 200 | 0.8048 | 0.8535 |
0.4723 | 0.78 | 300 | 0.3334 | 0.4693 |
0.3885 | 1.04 | 400 | 0.2105 | 0.2848 |
0.2986 | 1.3 | 500 | 0.1913 | 0.2544 |
0.2826 | 1.56 | 600 | 0.1811 | 0.2523 |
0.2613 | 1.82 | 700 | 0.1681 | 0.2505 |
0.2205 | 2.08 | 800 | 0.1530 | 0.2090 |
0.185 | 2.34 | 900 | 0.1427 | 0.2027 |
0.1947 | 2.6 | 1000 | 0.1471 | 0.2052 |
0.2077 | 2.86 | 1100 | 0.1425 | 0.2268 |
0.1724 | 3.12 | 1200 | 0.1466 | 0.2120 |
0.1674 | 3.39 | 1300 | 0.1420 | 0.1999 |
0.1531 | 3.65 | 1400 | 0.1637 | 0.2203 |
0.1596 | 3.91 | 1500 | 0.1429 | 0.1844 |
0.1322 | 4.17 | 1600 | 0.1228 | 0.1678 |
0.1351 | 4.43 | 1700 | 0.1223 | 0.1752 |
0.1301 | 4.69 | 1800 | 0.1171 | 0.1737 |
0.1264 | 4.95 | 1900 | 0.1237 | 0.1741 |
0.1182 | 5.21 | 2000 | 0.1166 | 0.1643 |
0.1121 | 5.47 | 2100 | 0.1259 | 0.1626 |
0.1152 | 5.73 | 2200 | 0.1222 | 0.1469 |
0.1153 | 5.99 | 2300 | 0.1262 | 0.1514 |
0.1009 | 6.25 | 2400 | 0.1369 | 0.1668 |
0.108 | 6.51 | 2500 | 0.1239 | 0.1605 |
0.0893 | 6.77 | 2600 | 0.1309 | 0.1423 |
0.1047 | 7.03 | 2700 | 0.1299 | 0.1520 |
0.0929 | 7.29 | 2800 | 0.1239 | 0.1582 |
0.0874 | 7.55 | 2900 | 0.1252 | 0.1466 |
0.0926 | 7.81 | 3000 | 0.1299 | 0.1459 |
0.0889 | 8.07 | 3100 | 0.1321 | 0.1353 |
0.0832 | 8.33 | 3200 | 0.1269 | 0.1319 |
0.0803 | 8.59 | 3300 | 0.1177 | 0.1308 |
0.072 | 8.85 | 3400 | 0.1160 | 0.1321 |
0.0718 | 9.11 | 3500 | 0.1375 | 0.1343 |
0.0727 | 9.38 | 3600 | 0.1504 | 0.1391 |
0.0639 | 9.64 | 3700 | 0.1284 | 0.1246 |
0.0632 | 9.9 | 3800 | 0.1280 | 0.1212 |
0.0633 | 10.16 | 3900 | 0.1278 | 0.1336 |
0.0632 | 10.42 | 4000 | 0.1241 | 0.1320 |
0.0595 | 10.68 | 4100 | 0.1321 | 0.1327 |
0.0625 | 10.94 | 4200 | 0.1318 | 0.1239 |
0.0581 | 11.2 | 4300 | 0.1397 | 0.1210 |
0.0568 | 11.46 | 4400 | 0.1418 | 0.1248 |
0.0576 | 11.72 | 4500 | 0.1242 | 0.1233 |
0.0533 | 11.98 | 4600 | 0.1370 | 0.1246 |
0.0469 | 12.24 | 4700 | 0.1358 | 0.1204 |
0.0512 | 12.5 | 4800 | 0.1386 | 0.1215 |
0.0493 | 12.76 | 4900 | 0.1464 | 0.1204 |
0.0522 | 13.02 | 5000 | 0.1411 | 0.1254 |
0.0512 | 13.28 | 5100 | 0.1478 | 0.1281 |
0.0442 | 13.54 | 5200 | 0.1282 | 0.1218 |
0.0435 | 13.8 | 5300 | 0.1520 | 0.1216 |
0.0404 | 14.06 | 5400 | 0.1428 | 0.1157 |
0.0425 | 14.32 | 5500 | 0.1409 | 0.1149 |
0.0384 | 14.58 | 5600 | 0.1429 | 0.1127 |
0.0366 | 14.84 | 5700 | 0.1398 | 0.1132 |
0.0398 | 15.1 | 5800 | 0.1362 | 0.1171 |
0.0374 | 15.36 | 5900 | 0.1420 | 0.1110 |
0.0342 | 15.62 | 6000 | 0.1403 | 0.1098 |
0.038 | 15.89 | 6100 | 0.1400 | 0.1082 |
0.0389 | 16.15 | 6200 | 0.1427 | 0.1100 |
0.034 | 16.41 | 6300 | 0.1498 | 0.1077 |
0.0317 | 16.67 | 6400 | 0.1478 | 0.1045 |
0.0321 | 16.93 | 6500 | 0.1463 | 0.1051 |
0.0305 | 17.19 | 6600 | 0.1520 | 0.1060 |
0.0327 | 17.45 | 6700 | 0.1414 | 0.1040 |
0.0302 | 17.71 | 6800 | 0.1539 | 0.1053 |
0.0334 | 17.97 | 6900 | 0.1494 | 0.1041 |
0.0288 | 18.23 | 7000 | 0.1517 | 0.1028 |
0.0274 | 18.49 | 7100 | 0.1522 | 0.1013 |
0.026 | 18.75 | 7200 | 0.1573 | 0.1016 |
0.03 | 19.01 | 7300 | 0.1531 | 0.1008 |
0.0258 | 19.27 | 7400 | 0.1547 | 0.0996 |
0.0254 | 19.53 | 7500 | 0.1545 | 0.1004 |
0.0242 | 19.79 | 7600 | 0.1542 | 0.0998 |
Framework versions
- Transformers 4.26.0
- Pytorch 1.10.0+cu113
- Datasets 2.9.0
- Tokenizers 0.13.2