metadata
license: cc-by-nc-4.0
base_model: nguyenvulebinh/wav2vec2-base-vietnamese-250h
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: >-
w2v2_ablation_focal_ctc_a0.75_g2.0-best_on-ling_head-tp0.025_tl10_fp0.001_fl16
results: []
w2v2_ablation_focal_ctc_a0.75_g2.0-best_on-ling_head-tp0.025_tl10_fp0.001_fl16
This model is a fine-tuned version of nguyenvulebinh/wav2vec2-base-vietnamese-250h on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.8829
- Wer: 0.0879
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: 2e-05
- train_batch_size: 8
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- total_train_batch_size: 32
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1337.3802 | 0.94 | 100 | 875.6535 | 18.6404 |
| 928.4498 | 1.89 | 200 | 336.8592 | 17.0854 |
| 159.8141 | 2.83 | 300 | 65.9143 | 1.0 |
| 84.4352 | 3.77 | 400 | 60.3730 | 1.0 |
| 77.6086 | 4.72 | 500 | 57.3593 | 1.0 |
| 74.6091 | 5.66 | 600 | 56.1616 | 1.0 |
| 73.5983 | 6.6 | 700 | 55.2774 | 1.0 |
| 72.9967 | 7.55 | 800 | 54.6511 | 1.0 |
| 71.2266 | 8.49 | 900 | 54.5362 | 1.0 |
| 69.7741 | 9.43 | 1000 | 51.8718 | 0.9648 |
| 58.1878 | 10.38 | 1100 | 28.9001 | 0.5655 |
| 32.9238 | 11.32 | 1200 | 12.7097 | 0.2391 |
| 21.0735 | 12.26 | 1300 | 8.5885 | 0.1785 |
| 15.9281 | 13.21 | 1400 | 6.8959 | 0.1529 |
| 13.7108 | 14.15 | 1500 | 5.7514 | 0.1392 |
| 11.2293 | 15.09 | 1600 | 4.9739 | 0.1244 |
| 10.3682 | 16.04 | 1700 | 4.5084 | 0.1237 |
| 9.6654 | 16.98 | 1800 | 4.3703 | 0.1259 |
| 8.816 | 17.92 | 1900 | 4.1278 | 0.1143 |
| 8.8608 | 18.87 | 2000 | 3.9105 | 0.1074 |
| 7.8629 | 19.81 | 2100 | 3.9114 | 0.1237 |
| 7.8569 | 20.75 | 2200 | 3.7354 | 0.1121 |
| 7.3392 | 21.7 | 2300 | 3.6668 | 0.1056 |
| 7.2164 | 22.64 | 2400 | 3.5747 | 0.1128 |
| 7.2758 | 23.58 | 2500 | 3.4933 | 0.1016 |
| 6.4516 | 24.53 | 2600 | 3.4821 | 0.0988 |
| 6.45 | 25.47 | 2700 | 3.3720 | 0.0996 |
| 6.0068 | 26.42 | 2800 | 3.4425 | 0.1044 |
| 5.5781 | 27.36 | 2900 | 3.3221 | 0.1014 |
| 5.5837 | 28.3 | 3000 | 3.4974 | 0.1041 |
| 5.7895 | 29.25 | 3100 | 3.3536 | 0.0950 |
| 5.6272 | 30.19 | 3200 | 3.2036 | 0.0960 |
| 5.594 | 31.13 | 3300 | 3.1747 | 0.0913 |
| 4.791 | 32.08 | 3400 | 3.1225 | 0.1038 |
| 5.0596 | 33.02 | 3500 | 3.2113 | 0.1095 |
| 4.985 | 33.96 | 3600 | 3.0622 | 0.0929 |
| 4.731 | 34.91 | 3700 | 3.0940 | 0.0956 |
| 4.6287 | 35.85 | 3800 | 3.0453 | 0.0961 |
| 4.5235 | 36.79 | 3900 | 3.0351 | 0.1019 |
| 4.7715 | 37.74 | 4000 | 3.0237 | 0.0928 |
| 4.7101 | 38.68 | 4100 | 3.0250 | 0.0943 |
| 4.243 | 39.62 | 4200 | 2.9704 | 0.0980 |
| 4.4015 | 40.57 | 4300 | 2.9600 | 0.0871 |
| 4.4545 | 41.51 | 4400 | 2.9806 | 0.0858 |
| 4.662 | 42.45 | 4500 | 2.9668 | 0.0969 |
| 4.0696 | 43.4 | 4600 | 2.9349 | 0.0935 |
| 3.5668 | 44.34 | 4700 | 2.9190 | 0.0917 |
| 3.8214 | 45.28 | 4800 | 2.9490 | 0.0901 |
| 3.8215 | 46.23 | 4900 | 2.9371 | 0.0912 |
| 3.6593 | 47.17 | 5000 | 2.9408 | 0.0875 |
| 3.3709 | 48.11 | 5100 | 2.9577 | 0.0920 |
| 3.5768 | 49.06 | 5200 | 2.9863 | 0.0940 |
| 3.3018 | 50.0 | 5300 | 2.9437 | 0.1003 |
| 3.2921 | 50.94 | 5400 | 2.9195 | 0.0923 |
| 3.4551 | 51.89 | 5500 | 2.9410 | 0.0950 |
| 3.6576 | 52.83 | 5600 | 2.9520 | 0.1011 |
| 3.5078 | 53.77 | 5700 | 2.8926 | 0.0937 |
| 3.0777 | 54.72 | 5800 | 2.8971 | 0.0913 |
| 3.0572 | 55.66 | 5900 | 2.8693 | 0.0891 |
| 3.0486 | 56.6 | 6000 | 2.8876 | 0.0882 |
| 3.1283 | 57.55 | 6100 | 2.8597 | 0.0913 |
| 2.8705 | 58.49 | 6200 | 2.9080 | 0.0904 |
| 3.0644 | 59.43 | 6300 | 2.9106 | 0.0917 |
| 2.8822 | 60.38 | 6400 | 2.9231 | 0.0891 |
| 3.2338 | 61.32 | 6500 | 2.9511 | 0.0903 |
| 3.048 | 62.26 | 6600 | 2.9539 | 0.0898 |
| 3.094 | 63.21 | 6700 | 2.9490 | 0.0908 |
| 3.0581 | 64.15 | 6800 | 2.8952 | 0.0886 |
| 2.9343 | 65.09 | 6900 | 2.8926 | 0.0883 |
| 2.9497 | 66.04 | 7000 | 2.8732 | 0.0888 |
| 2.7788 | 66.98 | 7100 | 2.8837 | 0.0904 |
| 2.7765 | 67.92 | 7200 | 2.9169 | 0.0951 |
| 3.134 | 68.87 | 7300 | 2.9030 | 0.0926 |
| 2.8812 | 69.81 | 7400 | 2.9045 | 0.0921 |
| 2.615 | 70.75 | 7500 | 2.9148 | 0.0871 |
| 2.5678 | 71.7 | 7600 | 2.9435 | 0.0922 |
| 2.4858 | 72.64 | 7700 | 2.9050 | 0.0928 |
| 2.5367 | 73.58 | 7800 | 2.8948 | 0.0878 |
| 2.3228 | 74.53 | 7900 | 2.8995 | 0.0891 |
| 2.5849 | 75.47 | 8000 | 2.9289 | 0.0928 |
| 2.6645 | 76.42 | 8100 | 2.8950 | 0.0884 |
| 2.6634 | 77.36 | 8200 | 2.9194 | 0.0922 |
| 2.393 | 78.3 | 8300 | 2.9074 | 0.0919 |
| 3.0675 | 79.25 | 8400 | 2.8927 | 0.0908 |
| 2.6344 | 80.19 | 8500 | 2.8768 | 0.0891 |
| 2.5742 | 81.13 | 8600 | 2.8809 | 0.0911 |
| 2.6523 | 82.08 | 8700 | 2.8639 | 0.0863 |
| 2.2657 | 83.02 | 8800 | 2.8809 | 0.0912 |
| 2.3238 | 83.96 | 8900 | 2.8764 | 0.0893 |
| 2.3664 | 84.91 | 9000 | 2.8738 | 0.0913 |
| 2.5655 | 85.85 | 9100 | 2.8876 | 0.0904 |
| 2.4372 | 86.79 | 9200 | 2.9024 | 0.0910 |
| 2.5267 | 87.74 | 9300 | 2.8922 | 0.0898 |
| 2.471 | 88.68 | 9400 | 2.8893 | 0.0884 |
| 2.5225 | 89.62 | 9500 | 2.8852 | 0.0888 |
| 2.4752 | 90.57 | 9600 | 2.8876 | 0.0892 |
| 2.5029 | 91.51 | 9700 | 2.8883 | 0.0885 |
| 2.7052 | 92.45 | 9800 | 2.8825 | 0.0871 |
| 2.4682 | 93.4 | 9900 | 2.8780 | 0.0870 |
| 2.3672 | 94.34 | 10000 | 2.8810 | 0.0872 |
| 2.5325 | 95.28 | 10100 | 2.8842 | 0.0884 |
| 2.4877 | 96.23 | 10200 | 2.8833 | 0.0884 |
| 2.7373 | 97.17 | 10300 | 2.8825 | 0.0882 |
| 2.5574 | 98.11 | 10400 | 2.8833 | 0.0881 |
| 2.2097 | 99.06 | 10500 | 2.8823 | 0.0883 |
| 2.5919 | 100.0 | 10600 | 2.8829 | 0.0879 |
Framework versions
- Transformers 4.35.2
- Pytorch 1.13.1+cu117
- Datasets 2.12.0
- Tokenizers 0.14.1