|
--- |
|
license: apache-2.0 |
|
base_model: facebook/wav2vec2-large-xlsr-53 |
|
tags: |
|
- automatic-speech-recognition |
|
- ./sample_speech.py |
|
- generated_from_trainer |
|
model-index: |
|
- name: ja-xlsr |
|
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. --> |
|
|
|
# ja-xlsr |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the ./SAMPLE_SPEECH.PY - NA dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 2.5952 |
|
- Cer: 0.3240 |
|
|
|
## 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: 4 |
|
- eval_batch_size: 4 |
|
- seed: 42 |
|
- distributed_type: multi-GPU |
|
- num_devices: 4 |
|
- total_train_batch_size: 16 |
|
- total_eval_batch_size: 16 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_steps: 50 |
|
- num_epochs: 300 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Cer | |
|
|:-------------:|:------:|:----:|:---------------:|:------:| |
|
| 4.9138 | 6.52 | 150 | 4.7965 | 1.0 | |
|
| 4.7484 | 13.04 | 300 | 4.6081 | 1.0 | |
|
| 4.5894 | 19.57 | 450 | 4.4697 | 0.9851 | |
|
| 4.2024 | 26.09 | 600 | 4.0373 | 0.9077 | |
|
| 2.7314 | 32.61 | 750 | 2.5507 | 0.5341 | |
|
| 1.2293 | 39.13 | 900 | 2.0146 | 0.4139 | |
|
| 0.5544 | 45.65 | 1050 | 1.9821 | 0.3556 | |
|
| 0.3224 | 52.17 | 1200 | 2.0190 | 0.3587 | |
|
| 0.1951 | 58.7 | 1350 | 2.1229 | 0.3612 | |
|
| 0.1539 | 65.22 | 1500 | 2.1114 | 0.3470 | |
|
| 0.1165 | 71.74 | 1650 | 2.2748 | 0.3315 | |
|
| 0.1119 | 78.26 | 1800 | 2.2391 | 0.3488 | |
|
| 0.0989 | 84.78 | 1950 | 2.3438 | 0.3383 | |
|
| 0.0915 | 91.3 | 2100 | 2.1218 | 0.3587 | |
|
| 0.0721 | 97.83 | 2250 | 2.2428 | 0.3519 | |
|
| 0.0742 | 104.35 | 2400 | 2.2293 | 0.3364 | |
|
| 0.0629 | 110.87 | 2550 | 2.2878 | 0.3371 | |
|
| 0.0495 | 117.39 | 2700 | 2.2672 | 0.3408 | |
|
| 0.0466 | 123.91 | 2850 | 2.2532 | 0.3525 | |
|
| 0.0424 | 130.43 | 3000 | 2.2844 | 0.3259 | |
|
| 0.0446 | 136.96 | 3150 | 2.2763 | 0.3253 | |
|
| 0.0411 | 143.48 | 3300 | 2.3011 | 0.3302 | |
|
| 0.0419 | 150.0 | 3450 | 2.3201 | 0.3420 | |
|
| 0.0333 | 156.52 | 3600 | 2.3644 | 0.3439 | |
|
| 0.0384 | 163.04 | 3750 | 2.3685 | 0.3532 | |
|
| 0.0367 | 169.57 | 3900 | 2.3970 | 0.3470 | |
|
| 0.0307 | 176.09 | 4050 | 2.3530 | 0.3309 | |
|
| 0.0328 | 182.61 | 4200 | 2.3415 | 0.3315 | |
|
| 0.0271 | 189.13 | 4350 | 2.4165 | 0.3309 | |
|
| 0.0213 | 195.65 | 4500 | 2.4478 | 0.3451 | |
|
| 0.0193 | 202.17 | 4650 | 2.5241 | 0.3556 | |
|
| 0.0204 | 208.7 | 4800 | 2.5700 | 0.3463 | |
|
| 0.0185 | 215.22 | 4950 | 2.5837 | 0.3178 | |
|
| 0.0161 | 221.74 | 5100 | 2.5139 | 0.3377 | |
|
| 0.0167 | 228.26 | 5250 | 2.5288 | 0.3352 | |
|
| 0.0148 | 234.78 | 5400 | 2.5741 | 0.3389 | |
|
| 0.0141 | 241.3 | 5550 | 2.5174 | 0.3389 | |
|
| 0.0122 | 247.83 | 5700 | 2.5573 | 0.3352 | |
|
| 0.0115 | 254.35 | 5850 | 2.5790 | 0.3296 | |
|
| 0.0141 | 260.87 | 6000 | 2.5774 | 0.3203 | |
|
| 0.0123 | 267.39 | 6150 | 2.6147 | 0.3309 | |
|
| 0.0214 | 273.91 | 6300 | 2.6202 | 0.3302 | |
|
| 0.0107 | 280.43 | 6450 | 2.6264 | 0.3234 | |
|
| 0.0086 | 286.96 | 6600 | 2.6075 | 0.3216 | |
|
| 0.0106 | 293.48 | 6750 | 2.5960 | 0.3247 | |
|
| 0.0085 | 300.0 | 6900 | 2.5952 | 0.3240 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.34.0 |
|
- Pytorch 2.1.0+cu121 |
|
- Datasets 2.14.5 |
|
- Tokenizers 0.14.1 |
|
|
