<!-- 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. -->
t5-base-TEDxJP-4front-1body-0rear
This model is a fine-tuned version of sonoisa/t5-base-japanese on the te_dx_jp dataset. It achieves the following results on the evaluation set:
- Loss: 0.4643
- Wer: 0.1751
- Mer: 0.1690
- Wil: 0.2562
- Wip: 0.7438
- Hits: 55598
- Substitutions: 6434
- Deletions: 2555
- Insertions: 2317
- Cer: 0.1374
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: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Mer | Wil | Wip | Hits | Substitutions | Deletions | Insertions | Cer |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6492 | 1.0 | 1457 | 0.4952 | 0.2272 | 0.2114 | 0.3015 | 0.6985 | 54739 | 6847 | 3001 | 4827 | 0.2013 |
| 0.5556 | 2.0 | 2914 | 0.4456 | 0.1899 | 0.1818 | 0.2686 | 0.7314 | 55189 | 6420 | 2978 | 2864 | 0.1558 |
| 0.4942 | 3.0 | 4371 | 0.4423 | 0.1814 | 0.1743 | 0.2614 | 0.7386 | 55493 | 6437 | 2657 | 2623 | 0.1457 |
| 0.4326 | 4.0 | 5828 | 0.4361 | 0.1749 | 0.1690 | 0.2561 | 0.7439 | 55542 | 6419 | 2626 | 2249 | 0.1362 |
| 0.3867 | 5.0 | 7285 | 0.4395 | 0.1752 | 0.1692 | 0.2559 | 0.7441 | 55542 | 6378 | 2667 | 2270 | 0.1374 |
| 0.3501 | 6.0 | 8742 | 0.4487 | 0.1751 | 0.1691 | 0.2565 | 0.7435 | 55598 | 6448 | 2541 | 2323 | 0.1366 |
| 0.3835 | 7.0 | 10199 | 0.4494 | 0.1744 | 0.1685 | 0.2556 | 0.7444 | 55594 | 6416 | 2577 | 2274 | 0.1378 |
| 0.3013 | 8.0 | 11656 | 0.4580 | 0.1744 | 0.1685 | 0.2563 | 0.7437 | 55570 | 6467 | 2550 | 2248 | 0.1366 |
| 0.3126 | 9.0 | 13113 | 0.4598 | 0.1749 | 0.1689 | 0.2564 | 0.7436 | 55571 | 6447 | 2569 | 2281 | 0.1376 |
| 0.3089 | 10.0 | 14570 | 0.4643 | 0.1751 | 0.1690 | 0.2562 | 0.7438 | 55598 | 6434 | 2555 | 2317 | 0.1374 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.12.1
