whisper-small-ko
This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3342
- Cer: 9.3562
- Wer: 23.5998
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: 1e-05
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 25000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Validation Loss | Step | Epoch | Cer | Wer |
|---|---|---|---|---|---|
| 0.8798 | 0.4962 | 500 | 0.2781 | 12.9471 | 32.0342 |
| 0.6216 | 0.4342 | 1000 | 0.5562 | 12.3301 | 30.0835 |
| 0.5545 | 0.4136 | 1500 | 0.8343 | 11.6154 | 28.6587 |
| 0.5038 | 0.3996 | 2000 | 1.1123 | 11.2510 | 27.8611 |
| 0.4938 | 0.3895 | 2500 | 1.3904 | 10.9614 | 27.1911 |
| 0.472 | 0.3840 | 3000 | 1.6685 | 10.7938 | 26.8655 |
| 0.46 | 0.3792 | 3500 | 1.9466 | 10.7587 | 26.7165 |
| 0.4475 | 0.3754 | 4000 | 2.2247 | 10.6994 | 26.5890 |
| 0.4396 | 0.3725 | 4500 | 2.5028 | 10.5686 | 26.3647 |
| 0.4335 | 0.3729 | 5000 | 2.7809 | 10.5336 | 26.2592 |
| 0.429 | 0.3726 | 5500 | 3.0590 | 10.4199 | 26.1158 |
| 0.424 | 0.3675 | 6000 | 3.3370 | 10.3781 | 25.9293 |
| 0.4191 | 0.3644 | 6500 | 3.6151 | 10.2905 | 25.7397 |
| 0.4115 | 0.3616 | 7000 | 3.8932 | 10.2256 | 25.5453 |
| 0.3951 | 0.3613 | 7500 | 4.1713 | 10.2008 | 25.4925 |
| 0.4018 | 0.3589 | 8000 | 4.4494 | 10.0662 | 25.2825 |
| 0.3957 | 0.3568 | 8500 | 4.7275 | 10.0739 | 25.2567 |
| 0.3944 | 0.3558 | 9000 | 5.0056 | 10.0189 | 25.1651 |
| 0.3879 | 0.3557 | 9500 | 5.2836 | 9.9843 | 25.1215 |
| 0.3849 | 0.3555 | 10000 | 5.5617 | 9.9827 | 25.0944 |
| 0.3875 | 0.3555 | 10500 | 5.8398 | 9.9046 | 25.0302 |
| 0.3787 | 0.3535 | 11000 | 6.1179 | 9.8685 | 24.8701 |
| 0.3751 | 0.3523 | 11500 | 6.3960 | 9.8697 | 24.8279 |
| 0.3765 | 0.3500 | 12000 | 6.6741 | 9.8013 | 24.6271 |
| 0.3718 | 0.3467 | 12500 | 6.9522 | 9.7820 | 24.5162 |
| 0.3574 | 0.3466 | 13000 | 7.2303 | 9.6958 | 24.4707 |
| 0.357 | 0.3439 | 13500 | 7.5083 | 9.6296 | 24.2872 |
| 0.3575 | 0.3420 | 14000 | 7.7864 | 9.6123 | 24.2847 |
| 0.3437 | 0.3414 | 14500 | 8.0645 | 9.5519 | 24.1108 |
| 0.3532 | 0.3410 | 15000 | 8.3426 | 9.4921 | 24.0301 |
| 0.345 | 0.3404 | 15500 | 8.6207 | 9.5022 | 23.9979 |
| 0.3456 | 0.3391 | 16000 | 8.8988 | 9.5080 | 23.9553 |
| 0.334 | 0.3394 | 16500 | 9.1769 | 9.4458 | 23.8505 |
| 0.3394 | 0.3378 | 17000 | 9.4549 | 9.4214 | 23.8336 |
| 0.3383 | 0.3371 | 17500 | 9.7330 | 9.4591 | 23.8312 |
| 0.3353 | 0.3370 | 18000 | 10.0111 | 9.3743 | 23.7401 |
| 0.3337 | 0.3366 | 18500 | 10.2892 | 9.3624 | 23.7240 |
| 0.336 | 0.3361 | 19000 | 10.5673 | 9.4365 | 23.7595 |
| 0.3287 | 0.3359 | 19500 | 10.8454 | 9.3547 | 23.6841 |
| 0.3266 | 0.3358 | 20000 | 11.1235 | 9.3734 | 23.6911 |
| 0.3316 | 0.3357 | 20500 | 11.4016 | 0.3357 | 23.6929 |
| 0.3314 | 0.3351 | 21000 | 11.6796 | 0.3351 | 23.6263 |
| 0.3315 | 0.3349 | 21500 | 11.9577 | 0.3349 | 23.5889 |
| 0.3268 | 0.3348 | 22000 | 12.2358 | 0.3348 | 23.5430 |
| 0.3263 | 0.3350 | 22500 | 12.5139 | 0.3350 | 23.5920 |
| 0.325 | 0.3342 | 23000 | 12.7920 | 0.3342 | 23.5998 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.1+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1
- Downloads last month
- 31
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.