<!-- 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-large-xls-r-300m-j-roman-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 1.2233
- Wer: 0.1437
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: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
4.3479 | 1.22 | 400 | 1.6349 | 0.3868 |
0.8621 | 2.45 | 800 | 1.0185 | 0.2346 |
0.5421 | 3.67 | 1200 | 0.7549 | 0.1868 |
0.3932 | 4.89 | 1600 | 0.7893 | 0.1811 |
0.332 | 6.12 | 2000 | 0.9318 | 0.1919 |
0.2902 | 7.34 | 2400 | 0.8263 | 0.1839 |
0.2542 | 8.56 | 2800 | 0.8491 | 0.1829 |
0.2355 | 9.79 | 3200 | 0.8820 | 0.1805 |
0.2206 | 11.01 | 3600 | 0.9183 | 0.1748 |
0.2041 | 12.23 | 4000 | 0.9131 | 0.1725 |
0.1878 | 13.46 | 4400 | 0.9075 | 0.1699 |
0.1733 | 14.68 | 4800 | 0.8456 | 0.1665 |
0.1746 | 15.9 | 5200 | 0.9353 | 0.1745 |
0.1671 | 17.13 | 5600 | 0.9318 | 0.1713 |
0.1641 | 18.35 | 6000 | 0.8804 | 0.1661 |
0.1578 | 19.57 | 6400 | 0.9849 | 0.1795 |
0.1534 | 20.8 | 6800 | 1.0036 | 0.1637 |
0.1484 | 22.02 | 7200 | 0.9618 | 0.1722 |
0.1431 | 23.24 | 7600 | 0.9947 | 0.1680 |
0.139 | 24.46 | 8000 | 0.9923 | 0.1729 |
0.134 | 25.69 | 8400 | 1.0015 | 0.1641 |
0.1298 | 26.91 | 8800 | 0.9930 | 0.1704 |
0.1253 | 28.13 | 9200 | 0.9977 | 0.1605 |
0.1178 | 29.36 | 9600 | 0.9756 | 0.1653 |
0.1178 | 30.58 | 10000 | 1.1122 | 0.1784 |
0.1165 | 31.8 | 10400 | 0.9883 | 0.1655 |
0.1073 | 33.03 | 10800 | 1.1286 | 0.1677 |
0.1121 | 34.25 | 11200 | 1.0406 | 0.1660 |
0.1081 | 35.47 | 11600 | 1.0976 | 0.1678 |
0.109 | 36.7 | 12000 | 1.0915 | 0.1722 |
0.1027 | 37.92 | 12400 | 1.1167 | 0.1712 |
0.0925 | 39.14 | 12800 | 1.1598 | 0.1693 |
0.0913 | 40.37 | 13200 | 1.0712 | 0.1640 |
0.0895 | 41.59 | 13600 | 1.1692 | 0.1745 |
0.0908 | 42.81 | 14000 | 1.1248 | 0.1641 |
0.0905 | 44.04 | 14400 | 1.0523 | 0.1678 |
0.0864 | 45.26 | 14800 | 1.0261 | 0.1626 |
0.0843 | 46.48 | 15200 | 1.0746 | 0.1676 |
0.0759 | 47.71 | 15600 | 1.1035 | 0.1596 |
0.0758 | 48.93 | 16000 | 1.0977 | 0.1622 |
0.0743 | 50.15 | 16400 | 1.1203 | 0.1677 |
0.0826 | 51.38 | 16800 | 1.0983 | 0.1651 |
0.0743 | 52.6 | 17200 | 1.1452 | 0.1622 |
0.0713 | 53.82 | 17600 | 1.0882 | 0.1623 |
0.0651 | 55.05 | 18000 | 1.0588 | 0.1608 |
0.0669 | 56.27 | 18400 | 1.1332 | 0.1600 |
0.0626 | 57.49 | 18800 | 1.0747 | 0.1562 |
0.0646 | 58.72 | 19200 | 1.0585 | 0.1599 |
0.0639 | 59.94 | 19600 | 1.0106 | 0.1543 |
0.0603 | 61.16 | 20000 | 1.0875 | 0.1585 |
0.0551 | 62.39 | 20400 | 1.1273 | 0.1537 |
0.0553 | 63.61 | 20800 | 1.1376 | 0.1577 |
0.052 | 64.83 | 21200 | 1.1429 | 0.1553 |
0.0506 | 66.06 | 21600 | 1.0872 | 0.1577 |
0.0495 | 67.28 | 22000 | 1.0954 | 0.1488 |
0.0483 | 68.5 | 22400 | 1.1397 | 0.1524 |
0.0421 | 69.72 | 22800 | 1.2144 | 0.1581 |
0.0457 | 70.95 | 23200 | 1.1581 | 0.1532 |
0.0405 | 72.17 | 23600 | 1.2150 | 0.1566 |
0.0409 | 73.39 | 24000 | 1.1176 | 0.1508 |
0.0386 | 74.62 | 24400 | 1.2018 | 0.1526 |
0.0374 | 75.84 | 24800 | 1.2548 | 0.1494 |
0.0376 | 77.06 | 25200 | 1.2161 | 0.1486 |
0.033 | 78.29 | 25600 | 1.1607 | 0.1558 |
0.0339 | 79.51 | 26000 | 1.1557 | 0.1498 |
0.0355 | 80.73 | 26400 | 1.1234 | 0.1490 |
0.031 | 81.96 | 26800 | 1.1778 | 0.1473 |
0.0301 | 83.18 | 27200 | 1.1594 | 0.1441 |
0.0292 | 84.4 | 27600 | 1.2036 | 0.1482 |
0.0256 | 85.63 | 28000 | 1.2334 | 0.1463 |
0.0259 | 86.85 | 28400 | 1.2072 | 0.1469 |
0.0271 | 88.07 | 28800 | 1.1843 | 0.1456 |
0.0241 | 89.3 | 29200 | 1.1712 | 0.1445 |
0.0223 | 90.52 | 29600 | 1.2059 | 0.1433 |
0.0213 | 91.74 | 30000 | 1.2231 | 0.1452 |
0.0212 | 92.97 | 30400 | 1.1980 | 0.1438 |
0.0223 | 94.19 | 30800 | 1.2148 | 0.1459 |
0.0185 | 95.41 | 31200 | 1.2190 | 0.1437 |
0.0202 | 96.64 | 31600 | 1.2051 | 0.1437 |
0.0188 | 97.86 | 32000 | 1.2154 | 0.1438 |
0.0183 | 99.08 | 32400 | 1.2233 | 0.1437 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1