|
--- |
|
license: apache-2.0 |
|
base_model: facebook/wav2vec2-xls-r-300m |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- audiofolder |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: wav2vec2-xls-r-300m-ja-syllable-cv-14 |
|
results: |
|
- task: |
|
name: Automatic Speech Recognition |
|
type: automatic-speech-recognition |
|
dataset: |
|
name: audiofolder |
|
type: audiofolder |
|
config: default |
|
split: train[:20%] |
|
args: default |
|
metrics: |
|
- name: Wer |
|
type: wer |
|
value: 0.04376879385232209 |
|
--- |
|
|
|
<!-- 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-xls-r-300m-ja-syllable-cv-14 |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co./facebook/wav2vec2-xls-r-300m) on Japanese using the train, dev, and validation splits of Common Voice 14.0. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.2005 |
|
- Wer: 0.0438 |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
Training: Common Voice 14.0 ja train, dev, validated |
|
Test: Common Voice 14.0 ja test[:20%] |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0001 |
|
- train_batch_size: 2 |
|
- eval_batch_size: 1 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 4 |
|
- total_train_batch_size: 8 |
|
- 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 | |
|
|:-------------:|:-----:|:------:|:---------------:|:------:| |
|
| 3.9103 | 0.37 | 2000 | 0.6143 | 0.1658 | |
|
| 0.5883 | 0.75 | 4000 | 0.4720 | 0.1340 | |
|
| 0.4759 | 1.12 | 6000 | 0.4080 | 0.1193 | |
|
| 0.4115 | 1.49 | 8000 | 0.3758 | 0.1173 | |
|
| 0.3833 | 1.87 | 10000 | 0.3591 | 0.1134 | |
|
| 0.3351 | 2.24 | 12000 | 0.3440 | 0.1011 | |
|
| 0.3129 | 2.61 | 14000 | 0.3550 | 0.1001 | |
|
| 0.3016 | 2.99 | 16000 | 0.3041 | 0.0949 | |
|
| 0.262 | 3.36 | 18000 | 0.2885 | 0.0853 | |
|
| 0.2571 | 3.73 | 20000 | 0.2825 | 0.0874 | |
|
| 0.2382 | 4.1 | 22000 | 0.2816 | 0.0848 | |
|
| 0.2171 | 4.48 | 24000 | 0.2732 | 0.0770 | |
|
| 0.2116 | 4.85 | 26000 | 0.2665 | 0.0773 | |
|
| 0.1964 | 5.22 | 28000 | 0.2703 | 0.0819 | |
|
| 0.1905 | 5.6 | 30000 | 0.2748 | 0.0822 | |
|
| 0.1855 | 5.97 | 32000 | 0.2572 | 0.0757 | |
|
| 0.1653 | 6.34 | 34000 | 0.2964 | 0.0803 | |
|
| 0.1684 | 6.72 | 36000 | 0.2744 | 0.0745 | |
|
| 0.1661 | 7.09 | 38000 | 0.2640 | 0.0790 | |
|
| 0.1504 | 7.46 | 40000 | 0.2803 | 0.0785 | |
|
| 0.1555 | 7.84 | 42000 | 0.2459 | 0.0703 | |
|
| 0.1408 | 8.21 | 44000 | 0.2666 | 0.0736 | |
|
| 0.1343 | 8.58 | 46000 | 0.2546 | 0.0711 | |
|
| 0.1358 | 8.96 | 48000 | 0.2582 | 0.0691 | |
|
| 0.1256 | 9.33 | 50000 | 0.2616 | 0.0709 | |
|
| 0.1245 | 9.7 | 52000 | 0.2621 | 0.0712 | |
|
| 0.1195 | 10.07 | 54000 | 0.2819 | 0.0692 | |
|
| 0.1122 | 10.45 | 56000 | 0.2666 | 0.0699 | |
|
| 0.1094 | 10.82 | 58000 | 0.2504 | 0.0666 | |
|
| 0.1062 | 11.19 | 60000 | 0.2610 | 0.0666 | |
|
| 0.1023 | 11.57 | 62000 | 0.2586 | 0.0656 | |
|
| 0.1036 | 11.94 | 64000 | 0.2463 | 0.0646 | |
|
| 0.096 | 12.31 | 66000 | 0.2677 | 0.0676 | |
|
| 0.0942 | 12.69 | 68000 | 0.2284 | 0.0607 | |
|
| 0.0939 | 13.06 | 70000 | 0.2663 | 0.0658 | |
|
| 0.0857 | 13.43 | 72000 | 0.2583 | 0.0653 | |
|
| 0.0889 | 13.81 | 74000 | 0.2215 | 0.0616 | |
|
| 0.0832 | 14.18 | 76000 | 0.2502 | 0.0631 | |
|
| 0.0813 | 14.55 | 78000 | 0.2472 | 0.0638 | |
|
| 0.0796 | 14.93 | 80000 | 0.2218 | 0.0600 | |
|
| 0.0774 | 15.3 | 82000 | 0.2376 | 0.0600 | |
|
| 0.0754 | 15.67 | 84000 | 0.2361 | 0.0588 | |
|
| 0.0745 | 16.04 | 86000 | 0.2578 | 0.0618 | |
|
| 0.0722 | 16.42 | 88000 | 0.2468 | 0.0604 | |
|
| 0.0709 | 16.79 | 90000 | 0.2268 | 0.0597 | |
|
| 0.0688 | 17.16 | 92000 | 0.2270 | 0.0555 | |
|
| 0.0665 | 17.54 | 94000 | 0.2320 | 0.0565 | |
|
| 0.0651 | 17.91 | 96000 | 0.2408 | 0.0600 | |
|
| 0.062 | 18.28 | 98000 | 0.2286 | 0.0550 | |
|
| 0.0609 | 18.66 | 100000 | 0.2314 | 0.0558 | |
|
| 0.0598 | 19.03 | 102000 | 0.2275 | 0.0547 | |
|
| 0.057 | 19.4 | 104000 | 0.2359 | 0.0547 | |
|
| 0.0559 | 19.78 | 106000 | 0.2501 | 0.0565 | |
|
| 0.0557 | 20.15 | 108000 | 0.2186 | 0.0530 | |
|
| 0.0519 | 20.52 | 110000 | 0.2281 | 0.0520 | |
|
| 0.0532 | 20.9 | 112000 | 0.2342 | 0.0525 | |
|
| 0.0521 | 21.27 | 114000 | 0.2265 | 0.0527 | |
|
| 0.0513 | 21.64 | 116000 | 0.2263 | 0.0528 | |
|
| 0.0485 | 22.01 | 118000 | 0.2343 | 0.0535 | |
|
| 0.0454 | 22.39 | 120000 | 0.2393 | 0.0517 | |
|
| 0.0454 | 22.76 | 122000 | 0.2314 | 0.0520 | |
|
| 0.0448 | 23.13 | 124000 | 0.2395 | 0.0493 | |
|
| 0.0444 | 23.51 | 126000 | 0.2299 | 0.0509 | |
|
| 0.0434 | 23.88 | 128000 | 0.2300 | 0.0499 | |
|
| 0.0402 | 24.25 | 130000 | 0.2314 | 0.0498 | |
|
| 0.0395 | 24.63 | 132000 | 0.2259 | 0.0478 | |
|
| 0.0383 | 25.0 | 134000 | 0.2202 | 0.0481 | |
|
| 0.0374 | 25.37 | 136000 | 0.2158 | 0.0484 | |
|
| 0.0375 | 25.75 | 138000 | 0.2165 | 0.0471 | |
|
| 0.0366 | 26.12 | 140000 | 0.2142 | 0.0469 | |
|
| 0.0347 | 26.49 | 142000 | 0.2139 | 0.0468 | |
|
| 0.0337 | 26.87 | 144000 | 0.2152 | 0.0477 | |
|
| 0.0343 | 27.24 | 146000 | 0.2059 | 0.0463 | |
|
| 0.0328 | 27.61 | 148000 | 0.2108 | 0.0469 | |
|
| 0.0324 | 27.99 | 150000 | 0.2061 | 0.0453 | |
|
| 0.0302 | 28.36 | 152000 | 0.2026 | 0.0450 | |
|
| 0.0316 | 28.73 | 154000 | 0.2057 | 0.0450 | |
|
| 0.0298 | 29.1 | 156000 | 0.2005 | 0.0439 | |
|
| 0.0301 | 29.48 | 158000 | 0.1983 | 0.0440 | |
|
| 0.0296 | 29.85 | 160000 | 0.2005 | 0.0438 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.31.0 |
|
- Pytorch 2.0.1+cu117 |
|
- Datasets 2.14.3 |
|
- Tokenizers 0.13.3 |
|
|
