<!-- 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-xlsr-mecita-coraa-portuguese-clean-grade-2-4
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.6426
- Wer: 0.9867
- Cer: 0.7546
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: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
30.5444 | 1.0 | 36 | 6.0803 | 1.0 | 1.0 |
30.5444 | 2.0 | 72 | 4.6965 | 1.0 | 1.0 |
10.2305 | 3.0 | 108 | 3.6812 | 1.0 | 1.0 |
10.2305 | 4.0 | 144 | 3.4078 | 0.965 | 0.9537 |
10.2305 | 5.0 | 180 | 3.2125 | 0.95 | 0.8998 |
4.437 | 6.0 | 216 | 3.1525 | 0.9492 | 0.9048 |
4.437 | 7.0 | 252 | 3.2147 | 0.9492 | 0.9146 |
4.437 | 8.0 | 288 | 3.0222 | 0.9508 | 0.8935 |
3.8259 | 9.0 | 324 | 2.9467 | 1.0 | 1.0 |
3.8259 | 10.0 | 360 | 2.9534 | 1.0 | 1.0 |
3.8259 | 11.0 | 396 | 2.8863 | 1.0 | 1.0 |
3.8025 | 12.0 | 432 | 2.8885 | 1.0 | 1.0 |
3.8025 | 13.0 | 468 | 2.8496 | 0.9992 | 0.9997 |
3.8906 | 14.0 | 504 | 2.8937 | 0.9967 | 0.9987 |
3.8906 | 15.0 | 540 | 2.8874 | 0.9717 | 0.9843 |
3.8906 | 16.0 | 576 | 2.8432 | 0.97 | 0.9739 |
3.9656 | 17.0 | 612 | 2.8498 | 0.9733 | 0.9747 |
3.9656 | 18.0 | 648 | 2.8246 | 0.9483 | 0.9380 |
3.9656 | 19.0 | 684 | 2.8476 | 0.955 | 0.9157 |
3.7122 | 20.0 | 720 | 2.8487 | 0.9608 | 0.9320 |
3.7122 | 21.0 | 756 | 2.9476 | 0.96 | 0.9413 |
3.7122 | 22.0 | 792 | 2.8280 | 0.9575 | 0.9146 |
3.4393 | 23.0 | 828 | 2.8241 | 0.9508 | 0.8927 |
3.4393 | 24.0 | 864 | 2.8278 | 0.965 | 0.9207 |
3.7183 | 25.0 | 900 | 2.8019 | 0.9575 | 0.9076 |
3.7183 | 26.0 | 936 | 2.7846 | 0.9525 | 0.8977 |
3.7183 | 27.0 | 972 | 2.8069 | 0.9642 | 0.8955 |
3.4876 | 28.0 | 1008 | 2.7941 | 0.96 | 0.8973 |
3.4876 | 29.0 | 1044 | 2.8143 | 0.9683 | 0.8711 |
3.4876 | 30.0 | 1080 | 2.7805 | 0.9683 | 0.8672 |
3.7973 | 31.0 | 1116 | 2.7645 | 0.9658 | 0.8532 |
3.7973 | 32.0 | 1152 | 2.7732 | 0.9742 | 0.8534 |
3.7973 | 33.0 | 1188 | 2.7968 | 0.9733 | 0.8537 |
3.4257 | 34.0 | 1224 | 2.7651 | 0.975 | 0.8587 |
3.4257 | 35.0 | 1260 | 2.7847 | 0.98 | 0.8449 |
3.4257 | 36.0 | 1296 | 2.7631 | 0.9817 | 0.8409 |
3.6149 | 37.0 | 1332 | 2.7716 | 0.9708 | 0.8436 |
3.6149 | 38.0 | 1368 | 2.7699 | 0.98 | 0.8302 |
3.5118 | 39.0 | 1404 | 2.7433 | 0.9792 | 0.8178 |
3.5118 | 40.0 | 1440 | 2.7654 | 0.9725 | 0.8181 |
3.5118 | 41.0 | 1476 | 2.7458 | 0.9817 | 0.8104 |
3.5272 | 42.0 | 1512 | 2.7577 | 0.9858 | 0.8121 |
3.5272 | 43.0 | 1548 | 2.7280 | 0.9933 | 0.8180 |
3.5272 | 44.0 | 1584 | 2.7223 | 0.9808 | 0.8140 |
3.3239 | 45.0 | 1620 | 2.7348 | 0.9842 | 0.8044 |
3.3239 | 46.0 | 1656 | 2.7225 | 0.9867 | 0.8044 |
3.3239 | 47.0 | 1692 | 2.7640 | 0.9875 | 0.7942 |
3.4254 | 48.0 | 1728 | 2.7388 | 0.9833 | 0.7955 |
3.4254 | 49.0 | 1764 | 2.7163 | 0.9867 | 0.7964 |
3.2168 | 50.0 | 1800 | 2.7176 | 0.99 | 0.7835 |
3.2168 | 51.0 | 1836 | 2.7010 | 0.9833 | 0.7896 |
3.2168 | 52.0 | 1872 | 2.7141 | 0.9867 | 0.7870 |
3.1638 | 53.0 | 1908 | 2.7013 | 0.985 | 0.7846 |
3.1638 | 54.0 | 1944 | 2.7287 | 0.9875 | 0.7917 |
3.1638 | 55.0 | 1980 | 2.6886 | 0.9892 | 0.7937 |
3.0805 | 56.0 | 2016 | 2.6875 | 0.9892 | 0.7793 |
3.0805 | 57.0 | 2052 | 2.7298 | 0.99 | 0.7876 |
3.0805 | 58.0 | 2088 | 2.7506 | 0.985 | 0.7829 |
3.1154 | 59.0 | 2124 | 2.6963 | 0.9892 | 0.7925 |
3.1154 | 60.0 | 2160 | 2.7002 | 0.9858 | 0.7823 |
3.1154 | 61.0 | 2196 | 2.6888 | 0.985 | 0.7819 |
2.9493 | 62.0 | 2232 | 2.7109 | 0.9825 | 0.7870 |
2.9493 | 63.0 | 2268 | 2.7069 | 0.9842 | 0.7780 |
2.8656 | 64.0 | 2304 | 2.7332 | 0.9842 | 0.7778 |
2.8656 | 65.0 | 2340 | 2.6759 | 0.9858 | 0.7841 |
2.8656 | 66.0 | 2376 | 2.6570 | 0.9858 | 0.7772 |
2.7412 | 67.0 | 2412 | 2.6872 | 0.9875 | 0.7659 |
2.7412 | 68.0 | 2448 | 2.7655 | 0.9817 | 0.7716 |
2.7412 | 69.0 | 2484 | 2.7470 | 0.98 | 0.7615 |
2.7649 | 70.0 | 2520 | 2.7192 | 0.9842 | 0.7736 |
2.7649 | 71.0 | 2556 | 2.6822 | 0.9792 | 0.7662 |
2.7649 | 72.0 | 2592 | 2.7063 | 0.9808 | 0.7671 |
2.7153 | 73.0 | 2628 | 2.7062 | 0.9792 | 0.7651 |
2.7153 | 74.0 | 2664 | 2.6431 | 0.9858 | 0.7706 |
2.6769 | 75.0 | 2700 | 2.6509 | 0.9892 | 0.7657 |
2.6769 | 76.0 | 2736 | 2.6543 | 0.985 | 0.7643 |
2.6769 | 77.0 | 2772 | 2.6779 | 0.9775 | 0.7646 |
2.6723 | 78.0 | 2808 | 2.6765 | 0.9833 | 0.7640 |
2.6723 | 79.0 | 2844 | 2.6687 | 0.985 | 0.7572 |
2.6723 | 80.0 | 2880 | 2.6857 | 0.9842 | 0.7634 |
2.7001 | 81.0 | 2916 | 2.6677 | 0.9858 | 0.7565 |
2.7001 | 82.0 | 2952 | 2.6569 | 0.9867 | 0.7538 |
2.7001 | 83.0 | 2988 | 2.6908 | 0.98 | 0.7538 |
2.6632 | 84.0 | 3024 | 2.6932 | 0.9858 | 0.7517 |
2.6632 | 85.0 | 3060 | 2.6426 | 0.9867 | 0.7546 |
2.6632 | 86.0 | 3096 | 2.6464 | 0.9825 | 0.7587 |
2.6488 | 87.0 | 3132 | 2.6865 | 0.9833 | 0.7602 |
2.6488 | 88.0 | 3168 | 2.6863 | 0.985 | 0.7577 |
2.7161 | 89.0 | 3204 | 2.6651 | 0.985 | 0.7546 |
2.7161 | 90.0 | 3240 | 2.6587 | 0.9825 | 0.7549 |
2.7161 | 91.0 | 3276 | 2.6709 | 0.9808 | 0.7552 |
2.6518 | 92.0 | 3312 | 2.6723 | 0.9883 | 0.7546 |
2.6518 | 93.0 | 3348 | 2.6523 | 0.985 | 0.7527 |
2.6518 | 94.0 | 3384 | 2.6503 | 0.9842 | 0.7514 |
2.6423 | 95.0 | 3420 | 2.6730 | 0.985 | 0.7544 |
2.6423 | 96.0 | 3456 | 2.6780 | 0.9917 | 0.7519 |
2.6423 | 97.0 | 3492 | 2.6833 | 0.9875 | 0.7547 |
2.762 | 98.0 | 3528 | 2.6739 | 0.9842 | 0.7509 |
2.762 | 99.0 | 3564 | 2.6718 | 0.9833 | 0.7511 |
2.6423 | 100.0 | 3600 | 2.6695 | 0.98 | 0.7530 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.5
- Tokenizers 0.13.3