xls-r-300m-et-children
This model is a fine-tuned version of TalTechNLP/xls-r-300m-et on an Estonian children's speech dataset.
More information about the model's performance and the data used for evaluation and training:
Luhtaru, Agnes; Jaaska, Rauno; Kruusamäe, Karl; Fishel, Mark (2023). Automatic Transcription for Estonian Children’s Speech. In: Proceedings of the 24th Nordic Conference on Computational Linguistics. https://openreview.net/forum?id=xbPTfBIUby
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- 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: 200
- num_epochs: 60
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.3485 | 1.61 | 100 | 0.2034 | 0.1782 |
0.1914 | 3.22 | 200 | 0.1818 | 0.1606 |
0.1641 | 4.83 | 300 | 0.1770 | 0.1572 |
0.1479 | 6.45 | 400 | 0.1745 | 0.1546 |
0.1374 | 8.06 | 500 | 0.1751 | 0.1538 |
0.1244 | 9.67 | 600 | 0.1734 | 0.1518 |
0.1211 | 11.29 | 700 | 0.1753 | 0.1508 |
0.1096 | 12.9 | 800 | 0.1758 | 0.1483 |
0.1059 | 14.51 | 900 | 0.1771 | 0.1469 |
0.0991 | 16.13 | 1000 | 0.1776 | 0.1469 |
0.0965 | 17.74 | 1100 | 0.1759 | 0.1469 |
0.0944 | 19.35 | 1200 | 0.1784 | 0.1459 |
0.0902 | 20.96 | 1300 | 0.1799 | 0.1469 |
0.0867 | 22.58 | 1400 | 0.1814 | 0.1440 |
0.0817 | 24.19 | 1500 | 0.1828 | 0.1438 |
0.0802 | 25.8 | 1600 | 0.1845 | 0.1438 |
0.0762 | 27.42 | 1700 | 0.1843 | 0.1431 |
0.0774 | 29.03 | 1800 | 0.1839 | 0.1432 |
0.0741 | 30.64 | 1900 | 0.1843 | 0.1442 |
0.0718 | 32.26 | 2000 | 0.1846 | 0.1429 |
0.07 | 33.86 | 2100 | 0.1852 | 0.1429 |
0.0692 | 35.48 | 2200 | 0.1872 | 0.1435 |
0.0671 | 37.1 | 2300 | 0.1874 | 0.1433 |
0.0667 | 38.7 | 2400 | 0.1887 | 0.1435 |
0.066 | 40.32 | 2500 | 0.1880 | 0.1422 |
0.0628 | 41.93 | 2600 | 0.1897 | 0.1426 |
0.0643 | 43.54 | 2700 | 0.1910 | 0.1428 |
0.0643 | 45.16 | 2800 | 0.1900 | 0.1431 |
0.0638 | 46.77 | 2900 | 0.1900 | 0.1427 |
0.0601 | 48.38 | 3000 | 0.1911 | 0.1431 |
0.0593 | 49.99 | 3100 | 0.1914 | 0.1432 |
0.0606 | 51.61 | 3200 | 0.1912 | 0.1433 |
0.0609 | 53.22 | 3300 | 0.1912 | 0.1431 |
0.0587 | 54.83 | 3400 | 0.1921 | 0.1429 |
0.0567 | 56.45 | 3500 | 0.1924 | 0.1430 |
0.0611 | 58.06 | 3600 | 0.1927 | 0.1431 |
0.0581 | 59.67 | 3700 | 0.1926 | 0.1430 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.12.1+rocm5.1.1
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2
- Downloads last month
- 1
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.