<!-- 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-10rear
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: 0.4646
- Wer: 0.1756
- Mer: 0.1698
- Wil: 0.2581
- Wip: 0.7419
- Hits: 55450
- Substitutions: 6518
- Deletions: 2619
- Insertions: 2202
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6456 | 1.0 | 1457 | 0.4955 | 0.2298 | 0.2127 | 0.3032 | 0.6968 | 54935 | 6936 | 2716 | 5190 | 0.2008 |
| 0.5333 | 2.0 | 2914 | 0.4472 | 0.1817 | 0.1755 | 0.2646 | 0.7354 | 55142 | 6584 | 2861 | 2291 | 0.1431 |
| 0.4864 | 3.0 | 4371 | 0.4420 | 0.1774 | 0.1714 | 0.2600 | 0.7400 | 55396 | 6542 | 2649 | 2266 | 0.1397 |
| 0.429 | 4.0 | 5828 | 0.4374 | 0.1764 | 0.1704 | 0.2587 | 0.7413 | 55446 | 6512 | 2629 | 2249 | 0.1389 |
| 0.3741 | 5.0 | 7285 | 0.4413 | 0.1744 | 0.1687 | 0.2559 | 0.7441 | 55518 | 6416 | 2653 | 2198 | 0.1383 |
| 0.3467 | 6.0 | 8742 | 0.4467 | 0.1742 | 0.1686 | 0.2564 | 0.7436 | 55493 | 6466 | 2628 | 2159 | 0.1390 |
| 0.3761 | 7.0 | 10199 | 0.4524 | 0.1754 | 0.1696 | 0.2577 | 0.7423 | 55471 | 6498 | 2618 | 2210 | 0.1380 |
| 0.3102 | 8.0 | 11656 | 0.4557 | 0.1751 | 0.1695 | 0.2574 | 0.7426 | 55412 | 6478 | 2697 | 2133 | 0.1395 |
| 0.3008 | 9.0 | 13113 | 0.4632 | 0.1758 | 0.1700 | 0.2584 | 0.7416 | 55421 | 6516 | 2650 | 2189 | 0.1387 |
| 0.3051 | 10.0 | 14570 | 0.4646 | 0.1756 | 0.1698 | 0.2581 | 0.7419 | 55450 | 6518 | 2619 | 2202 | 0.1383 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu116
- Datasets 2.4.0
- Tokenizers 0.12.1
