|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- timit_asr |
|
model-index: |
|
- name: test_lai_phonomes_transf |
|
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. --> |
|
|
|
# test_lai_phonomes_transf |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the timit_asr dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.3317 |
|
- Cer: 0.1174 |
|
|
|
## 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: 16 |
|
- 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 |
|
- training_steps: 2500 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Cer | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:| |
|
| 5.9718 | 0.4 | 100 | 3.5323 | 0.7939 | |
|
| 2.9356 | 0.8 | 200 | 2.1987 | 0.6826 | |
|
| 1.3018 | 1.2 | 300 | 1.1248 | 0.3547 | |
|
| 0.6407 | 1.61 | 400 | 0.4636 | 0.1714 | |
|
| 0.514 | 2.01 | 500 | 0.3942 | 0.1603 | |
|
| 0.4476 | 2.41 | 600 | 0.3745 | 0.1481 | |
|
| 0.4342 | 2.81 | 700 | 0.3387 | 0.1375 | |
|
| 0.3979 | 3.21 | 800 | 0.3433 | 0.1379 | |
|
| 0.4003 | 3.61 | 900 | 0.3596 | 0.1329 | |
|
| 0.3826 | 4.02 | 1000 | 0.3226 | 0.1322 | |
|
| 0.3487 | 4.42 | 1100 | 0.3338 | 0.1264 | |
|
| 0.338 | 4.82 | 1200 | 0.3159 | 0.1274 | |
|
| 0.3141 | 5.22 | 1300 | 0.3248 | 0.1257 | |
|
| 0.3011 | 5.62 | 1400 | 0.3363 | 0.1247 | |
|
| 0.2853 | 6.02 | 1500 | 0.3099 | 0.1215 | |
|
| 0.2436 | 6.43 | 1600 | 0.3113 | 0.1206 | |
|
| 0.253 | 6.83 | 1700 | 0.3054 | 0.1202 | |
|
| 0.236 | 7.23 | 1800 | 0.3369 | 0.1230 | |
|
| 0.2132 | 7.63 | 1900 | 0.3263 | 0.1190 | |
|
| 0.2179 | 8.03 | 2000 | 0.3195 | 0.1191 | |
|
| 0.1953 | 8.43 | 2100 | 0.3214 | 0.1189 | |
|
| 0.1855 | 8.84 | 2200 | 0.3285 | 0.1181 | |
|
| 0.1831 | 9.24 | 2300 | 0.3344 | 0.1179 | |
|
| 0.1714 | 9.64 | 2400 | 0.3363 | 0.1182 | |
|
| 0.1642 | 10.04 | 2500 | 0.3317 | 0.1174 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.17.0 |
|
- Pytorch 2.4.0 |
|
- Datasets 1.18.3 |
|
- Tokenizers 0.21.0 |
|
|
