<!-- 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-0front-1body-8rear
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: 0.4672
- Wer: 0.1759
- Mer: 0.1698
- Wil: 0.2574
- Wip: 0.7426
- Hits: 55537
- Substitutions: 6457
- Deletions: 2593
- Insertions: 2312
- Cer: 0.1383
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6417 | 1.0 | 1457 | 0.4928 | 0.2086 | 0.1973 | 0.2873 | 0.7127 | 54805 | 6751 | 3031 | 3693 | 0.1746 |
| 0.5435 | 2.0 | 2914 | 0.4511 | 0.1814 | 0.1751 | 0.2634 | 0.7366 | 55192 | 6518 | 2877 | 2322 | 0.1452 |
| 0.4914 | 3.0 | 4371 | 0.4424 | 0.1762 | 0.1704 | 0.2572 | 0.7428 | 55389 | 6383 | 2815 | 2180 | 0.1390 |
| 0.427 | 4.0 | 5828 | 0.4388 | 0.1751 | 0.1695 | 0.2569 | 0.7431 | 55408 | 6431 | 2748 | 2129 | 0.1366 |
| 0.3762 | 5.0 | 7285 | 0.4465 | 0.1747 | 0.1689 | 0.2561 | 0.7439 | 55533 | 6424 | 2630 | 2230 | 0.1361 |
| 0.3562 | 6.0 | 8742 | 0.4505 | 0.1761 | 0.1700 | 0.2581 | 0.7419 | 55558 | 6507 | 2522 | 2348 | 0.1402 |
| 0.3884 | 7.0 | 10199 | 0.4550 | 0.1750 | 0.1691 | 0.2564 | 0.7436 | 55548 | 6439 | 2600 | 2264 | 0.1364 |
| 0.3144 | 8.0 | 11656 | 0.4616 | 0.1760 | 0.1698 | 0.2572 | 0.7428 | 55571 | 6447 | 2569 | 2352 | 0.1373 |
| 0.3075 | 9.0 | 13113 | 0.4660 | 0.1761 | 0.1700 | 0.2572 | 0.7428 | 55547 | 6431 | 2609 | 2336 | 0.1400 |
| 0.3152 | 10.0 | 14570 | 0.4672 | 0.1759 | 0.1698 | 0.2574 | 0.7426 | 55537 | 6457 | 2593 | 2312 | 0.1383 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.12.1
