|
--- |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: Model_G_ALL_Wav2Vec2 |
|
results: [] |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# Model_G_ALL_Wav2Vec2 |
|
|
|
This model was trained from scratch on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.7956 |
|
- Wer: 0.1972 |
|
- Cer: 0.0811 |
|
|
|
## 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: 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 |
|
- lr_scheduler_warmup_steps: 500 |
|
- num_epochs: 30 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |
|
|:-------------:|:-----:|:-----:|:---------------:|:------:|:------:| |
|
| 0.8395 | 0.67 | 400 | 0.5656 | 0.3303 | 0.1236 | |
|
| 0.3196 | 1.34 | 800 | 0.5007 | 0.2922 | 0.1071 | |
|
| 0.2491 | 2.01 | 1200 | 0.5008 | 0.2830 | 0.1056 | |
|
| 0.2012 | 2.68 | 1600 | 0.5177 | 0.2689 | 0.1003 | |
|
| 0.1882 | 3.35 | 2000 | 0.5517 | 0.2622 | 0.0986 | |
|
| 0.1811 | 4.02 | 2400 | 0.5225 | 0.2543 | 0.0980 | |
|
| 0.1648 | 4.69 | 2800 | 0.5504 | 0.2477 | 0.0948 | |
|
| 0.1451 | 5.36 | 3200 | 0.5181 | 0.2346 | 0.0908 | |
|
| 0.149 | 6.04 | 3600 | 0.5204 | 0.2375 | 0.0941 | |
|
| 0.1328 | 6.71 | 4000 | 0.5780 | 0.2375 | 0.0920 | |
|
| 0.1256 | 7.38 | 4400 | 0.5406 | 0.2443 | 0.0960 | |
|
| 0.1193 | 8.05 | 4800 | 0.5489 | 0.2317 | 0.0928 | |
|
| 0.1099 | 8.72 | 5200 | 0.5864 | 0.2363 | 0.0925 | |
|
| 0.1125 | 9.39 | 5600 | 0.5749 | 0.2267 | 0.0902 | |
|
| 0.1044 | 10.06 | 6000 | 0.5698 | 0.2279 | 0.0905 | |
|
| 0.0925 | 10.73 | 6400 | 0.6051 | 0.2337 | 0.0933 | |
|
| 0.0951 | 11.4 | 6800 | 0.6785 | 0.2286 | 0.0907 | |
|
| 0.0926 | 12.07 | 7200 | 0.5937 | 0.2337 | 0.0919 | |
|
| 0.0838 | 12.74 | 7600 | 0.5918 | 0.2233 | 0.0893 | |
|
| 0.0775 | 13.41 | 8000 | 0.5642 | 0.2227 | 0.0888 | |
|
| 0.0742 | 14.08 | 8400 | 0.5927 | 0.2249 | 0.0898 | |
|
| 0.0687 | 14.75 | 8800 | 0.6647 | 0.2265 | 0.0900 | |
|
| 0.0685 | 15.42 | 9200 | 0.7438 | 0.2164 | 0.0885 | |
|
| 0.0645 | 16.09 | 9600 | 0.6351 | 0.2128 | 0.0858 | |
|
| 0.0582 | 16.76 | 10000 | 0.6164 | 0.2169 | 0.0878 | |
|
| 0.0604 | 17.44 | 10400 | 0.6327 | 0.2146 | 0.0867 | |
|
| 0.0557 | 18.11 | 10800 | 0.6790 | 0.2148 | 0.0879 | |
|
| 0.0552 | 18.78 | 11200 | 0.6859 | 0.2101 | 0.0848 | |
|
| 0.0474 | 19.45 | 11600 | 0.6648 | 0.2071 | 0.0847 | |
|
| 0.048 | 20.12 | 12000 | 0.7172 | 0.2136 | 0.0873 | |
|
| 0.0475 | 20.79 | 12400 | 0.6451 | 0.2058 | 0.0845 | |
|
| 0.041 | 21.46 | 12800 | 0.6826 | 0.2074 | 0.0839 | |
|
| 0.0405 | 22.13 | 13200 | 0.6738 | 0.2110 | 0.0842 | |
|
| 0.0355 | 22.8 | 13600 | 0.7020 | 0.2050 | 0.0839 | |
|
| 0.0325 | 23.47 | 14000 | 0.7085 | 0.2117 | 0.0854 | |
|
| 0.0308 | 24.14 | 14400 | 0.7418 | 0.2077 | 0.0854 | |
|
| 0.0321 | 24.81 | 14800 | 0.7371 | 0.2051 | 0.0840 | |
|
| 0.0274 | 25.48 | 15200 | 0.7611 | 0.2082 | 0.0848 | |
|
| 0.0287 | 26.15 | 15600 | 0.7208 | 0.2021 | 0.0836 | |
|
| 0.0253 | 26.82 | 16000 | 0.7432 | 0.2025 | 0.0831 | |
|
| 0.0256 | 27.49 | 16400 | 0.7435 | 0.2011 | 0.0824 | |
|
| 0.0243 | 28.16 | 16800 | 0.7543 | 0.1991 | 0.0818 | |
|
| 0.0241 | 28.83 | 17200 | 0.7676 | 0.1986 | 0.0814 | |
|
| 0.0204 | 29.51 | 17600 | 0.7956 | 0.1972 | 0.0811 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.31.0 |
|
- Pytorch 2.0.1+cu117 |
|
- Datasets 1.18.3 |
|
- Tokenizers 0.13.3 |
|
|
