|
--- |
|
license: apache-2.0 |
|
tags: |
|
- automatic-speech-recognition |
|
- hts98/original_ver1.2 |
|
- generated_from_trainer |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: wav2vec2-ver2.0 |
|
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. --> |
|
|
|
# wav2vec2-ver2.0 |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the HTS98/ORIGINAL_VER1.2 - NA dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.9210 |
|
- Wer: 0.3901 |
|
|
|
## 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: 5e-05 |
|
- train_batch_size: 8 |
|
- eval_batch_size: 16 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 4 |
|
- 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: 200 |
|
- num_epochs: 50.0 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:| |
|
| No log | 0.99 | 104 | 4.0281 | 1.0 | |
|
| No log | 2.0 | 209 | 3.4058 | 1.0 | |
|
| No log | 2.99 | 312 | 3.3457 | 1.0 | |
|
| No log | 4.0 | 417 | 3.3384 | 1.0 | |
|
| 3.351 | 5.0 | 522 | 3.3195 | 1.0 | |
|
| 3.351 | 6.0 | 627 | 3.2774 | 1.0000 | |
|
| 3.351 | 6.99 | 731 | 2.7498 | 1.0184 | |
|
| 3.351 | 8.0 | 836 | 2.0695 | 0.9268 | |
|
| 3.351 | 9.0 | 941 | 1.6666 | 0.8103 | |
|
| 2.5482 | 10.0 | 1046 | 1.4119 | 0.7014 | |
|
| 2.5482 | 10.99 | 1150 | 1.2480 | 0.6408 | |
|
| 2.5482 | 12.0 | 1255 | 1.1397 | 0.5814 | |
|
| 2.5482 | 13.0 | 1360 | 1.0593 | 0.5381 | |
|
| 2.5482 | 14.0 | 1465 | 1.0060 | 0.5099 | |
|
| 1.1172 | 14.99 | 1569 | 0.9678 | 0.4830 | |
|
| 1.1172 | 16.0 | 1674 | 0.9379 | 0.4692 | |
|
| 1.1172 | 17.0 | 1779 | 0.9127 | 0.4618 | |
|
| 1.1172 | 18.0 | 1884 | 0.8923 | 0.4352 | |
|
| 1.1172 | 18.99 | 1988 | 0.8827 | 0.4254 | |
|
| 0.7161 | 20.0 | 2093 | 0.8722 | 0.4304 | |
|
| 0.7161 | 21.0 | 2198 | 0.8755 | 0.4142 | |
|
| 0.7161 | 22.0 | 2303 | 0.8680 | 0.4157 | |
|
| 0.7161 | 22.99 | 2407 | 0.8705 | 0.4116 | |
|
| 0.5338 | 24.0 | 2512 | 0.8611 | 0.4039 | |
|
| 0.5338 | 25.0 | 2617 | 0.8716 | 0.3995 | |
|
| 0.5338 | 26.0 | 2722 | 0.8721 | 0.4037 | |
|
| 0.5338 | 26.99 | 2826 | 0.8809 | 0.3973 | |
|
| 0.5338 | 28.0 | 2931 | 0.9037 | 0.3938 | |
|
| 0.4299 | 29.0 | 3036 | 0.9119 | 0.3903 | |
|
| 0.4299 | 30.0 | 3141 | 0.9117 | 0.3912 | |
|
| 0.4299 | 30.99 | 3245 | 0.9027 | 0.3930 | |
|
| 0.4299 | 31.99 | 3328 | 0.9240 | 0.3898 | |
|
| 0.4299 | 33.0 | 3433 | 0.9337 | 0.3872 | |
|
| 0.3491 | 34.0 | 3538 | 0.9210 | 0.3901 | |
|
| 0.3491 | 35.0 | 3643 | 0.9309 | 0.3905 | |
|
| 0.3491 | 35.99 | 3747 | 0.9528 | 0.3906 | |
|
| 0.3491 | 37.0 | 3852 | 0.9506 | 0.3880 | |
|
| 0.3491 | 38.0 | 3957 | 0.9607 | 0.3853 | |
|
| 0.3195 | 39.0 | 4062 | 0.9567 | 0.3906 | |
|
| 0.3195 | 39.99 | 4166 | 0.9632 | 0.3893 | |
|
| 0.3195 | 41.0 | 4271 | 0.9797 | 0.3839 | |
|
| 0.3195 | 42.0 | 4376 | 0.9819 | 0.3854 | |
|
| 0.3195 | 43.0 | 4481 | 0.9686 | 0.3870 | |
|
| 0.2892 | 43.99 | 4585 | 0.9808 | 0.3895 | |
|
| 0.2892 | 45.0 | 4690 | 0.9857 | 0.3892 | |
|
| 0.2892 | 46.0 | 4795 | 0.9959 | 0.3831 | |
|
| 0.2892 | 47.0 | 4900 | 0.9959 | 0.3870 | |
|
| 0.2705 | 47.99 | 5004 | 1.0028 | 0.3860 | |
|
| 0.2705 | 49.0 | 5109 | 1.0019 | 0.3869 | |
|
| 0.2705 | 49.86 | 5200 | 1.0050 | 0.3857 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.31.0.dev0 |
|
- Pytorch 2.0.1+cu118 |
|
- Datasets 2.7.0 |
|
- Tokenizers 0.13.3 |
|
|