<!-- 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. -->
exp18-M03-both
This model is a fine-tuned version of yongjian/wav2vec2-large-a on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4134
- Wer: 0.8533
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: 0.0001
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 44.7613 | 0.35 | 500 | 3.3891 | 1.0525 |
| 3.1954 | 0.7 | 1000 | 2.8766 | 1.0 |
| 2.9522 | 1.05 | 1500 | 2.7578 | 1.0 |
| 2.843 | 1.4 | 2000 | 2.5628 | 1.1318 |
| 2.6645 | 1.75 | 2500 | 2.1406 | 1.2864 |
| 2.3587 | 2.1 | 3000 | 1.8164 | 1.2934 |
| 2.1731 | 2.45 | 3500 | 1.5732 | 1.2775 |
| 2.0242 | 2.8 | 4000 | 1.4249 | 1.2666 |
| 1.9453 | 3.15 | 4500 | 1.3079 | 1.2220 |
| 1.7871 | 3.5 | 5000 | 1.2389 | 1.2081 |
| 1.7147 | 3.85 | 5500 | 1.1724 | 1.2101 |
| 1.5729 | 4.2 | 6000 | 1.1638 | 1.1982 |
| 1.4966 | 4.55 | 6500 | 1.0529 | 1.1497 |
| 1.3898 | 4.9 | 7000 | 1.0808 | 1.1506 |
| 1.3447 | 5.25 | 7500 | 0.9702 | 1.1229 |
| 1.2342 | 5.6 | 8000 | 0.8994 | 1.1219 |
| 1.1918 | 5.95 | 8500 | 0.9212 | 1.1169 |
| 1.1037 | 6.3 | 9000 | 0.9057 | 1.1080 |
| 1.0661 | 6.65 | 9500 | 0.8231 | 1.1110 |
| 1.0501 | 7.0 | 10000 | 0.8291 | 1.0912 |
| 0.9069 | 7.35 | 10500 | 0.8360 | 1.0902 |
| 0.8959 | 7.7 | 11000 | 0.7961 | 1.0684 |
| 0.9256 | 8.05 | 11500 | 0.7459 | 1.0684 |
| 0.8686 | 8.4 | 12000 | 0.7276 | 1.0456 |
| 0.7998 | 8.75 | 12500 | 0.7195 | 1.0525 |
| 0.7406 | 9.1 | 13000 | 0.7471 | 1.0515 |
| 0.7646 | 9.45 | 13500 | 0.7716 | 1.0624 |
| 0.7018 | 9.8 | 14000 | 0.7262 | 1.0446 |
| 0.7114 | 10.15 | 14500 | 0.6795 | 1.0327 |
| 0.6498 | 10.5 | 15000 | 0.6724 | 1.0347 |
| 0.6652 | 10.85 | 15500 | 0.6994 | 1.0347 |
| 0.638 | 11.2 | 16000 | 0.6565 | 1.0159 |
| 0.6078 | 11.55 | 16500 | 0.6695 | 1.0575 |
| 0.588 | 11.9 | 17000 | 0.6391 | 1.0149 |
| 0.5722 | 12.25 | 17500 | 0.6321 | 1.0188 |
| 0.5505 | 12.6 | 18000 | 0.6306 | 1.0089 |
| 0.5297 | 12.95 | 18500 | 0.6100 | 1.0139 |
| 0.5188 | 13.3 | 19000 | 0.5426 | 0.9931 |
| 0.4865 | 13.65 | 19500 | 0.5410 | 0.9881 |
| 0.5132 | 14.0 | 20000 | 0.5095 | 0.9792 |
| 0.4782 | 14.35 | 20500 | 0.4962 | 0.9901 |
| 0.4627 | 14.7 | 21000 | 0.5277 | 0.9871 |
| 0.4568 | 15.05 | 21500 | 0.4958 | 0.9683 |
| 0.4312 | 15.4 | 22000 | 0.5146 | 0.9752 |
| 0.4286 | 15.75 | 22500 | 0.4682 | 0.9693 |
| 0.428 | 16.1 | 23000 | 0.5121 | 0.9851 |
| 0.3656 | 16.45 | 23500 | 0.4894 | 0.9485 |
| 0.3884 | 16.79 | 24000 | 0.4832 | 0.9465 |
| 0.3835 | 17.14 | 24500 | 0.4925 | 0.9841 |
| 0.3584 | 17.49 | 25000 | 0.5503 | 0.9782 |
| 0.3719 | 17.84 | 25500 | 0.4960 | 0.9415 |
| 0.3555 | 18.19 | 26000 | 0.4238 | 0.9594 |
| 0.3196 | 18.54 | 26500 | 0.4501 | 0.9495 |
| 0.3288 | 18.89 | 27000 | 0.5292 | 0.9564 |
| 0.3402 | 19.24 | 27500 | 0.4156 | 0.9475 |
| 0.2889 | 19.59 | 28000 | 0.4056 | 0.9633 |
| 0.3562 | 19.94 | 28500 | 0.3972 | 0.9504 |
| 0.336 | 20.29 | 29000 | 0.4021 | 0.9257 |
| 0.2952 | 20.64 | 29500 | 0.3920 | 0.9167 |
| 0.2678 | 20.99 | 30000 | 0.3610 | 0.9049 |
| 0.2816 | 21.34 | 30500 | 0.3782 | 0.9267 |
| 0.2718 | 21.69 | 31000 | 0.3502 | 0.9068 |
| 0.2948 | 22.04 | 31500 | 0.3412 | 0.9078 |
| 0.2782 | 22.39 | 32000 | 0.3799 | 0.9039 |
| 0.2668 | 22.74 | 32500 | 0.3725 | 0.9058 |
| 0.2685 | 23.09 | 33000 | 0.3825 | 0.8880 |
| 0.2514 | 23.44 | 33500 | 0.3618 | 0.8791 |
| 0.2305 | 23.79 | 34000 | 0.4211 | 0.8870 |
| 0.2671 | 24.14 | 34500 | 0.4126 | 0.8900 |
| 0.2153 | 24.49 | 35000 | 0.4106 | 0.8801 |
| 0.2323 | 24.84 | 35500 | 0.3845 | 0.8751 |
| 0.2208 | 25.19 | 36000 | 0.4017 | 0.8741 |
| 0.2023 | 25.54 | 36500 | 0.4451 | 0.8662 |
| 0.232 | 25.89 | 37000 | 0.4133 | 0.8583 |
| 0.2101 | 26.24 | 37500 | 0.4118 | 0.8662 |
| 0.2139 | 26.59 | 38000 | 0.3937 | 0.8682 |
| 0.1917 | 26.94 | 38500 | 0.4015 | 0.8603 |
| 0.1904 | 27.29 | 39000 | 0.4018 | 0.8622 |
| 0.2265 | 27.64 | 39500 | 0.3983 | 0.8573 |
| 0.2081 | 27.99 | 40000 | 0.4027 | 0.8563 |
| 0.2124 | 28.34 | 40500 | 0.4172 | 0.8523 |
| 0.191 | 28.69 | 41000 | 0.4018 | 0.8444 |
| 0.1906 | 29.04 | 41500 | 0.4148 | 0.8494 |
| 0.1613 | 29.39 | 42000 | 0.4195 | 0.8543 |
| 0.1864 | 29.74 | 42500 | 0.4134 | 0.8533 |
Framework versions
- Transformers 4.23.1
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.13.2
