Padomin's picture
update model card README.md
f7669c7
|
raw
history blame
3.1 kB
metadata
license: cc-by-sa-4.0
tags:
  - generated_from_trainer
datasets:
  - te_dx_jp
model-index:
  - name: t5-base-TEDxJP-6front-1body-6rear
    results: []

t5-base-TEDxJP-6front-1body-6rear

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: 0.4380
  • Wer: 0.1700
  • Mer: 0.1642
  • Wil: 0.2501
  • Wip: 0.7499
  • Hits: 55894
  • Substitutions: 6327
  • Deletions: 2366
  • Insertions: 2286
  • Cer: 0.1345

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
0.5938 1.0 1457 0.4764 0.2123 0.1997 0.2886 0.7114 54961 6701 2925 4085 0.1721
0.4817 2.0 2914 0.4166 0.1827 0.1754 0.2615 0.7385 55462 6356 2769 2676 0.1470
0.4467 3.0 4371 0.4119 0.1715 0.1660 0.2530 0.7470 55677 6410 2500 2169 0.1339
0.3818 4.0 5828 0.4134 0.1714 0.1654 0.2522 0.7478 55837 6396 2354 2319 0.1340
0.3577 5.0 7285 0.4171 0.1716 0.1653 0.2509 0.7491 55938 6303 2346 2432 0.1339
0.3222 6.0 8742 0.4195 0.1681 0.1628 0.2484 0.7516 55829 6282 2476 2099 0.1314
0.2938 7.0 10199 0.4242 0.1685 0.1634 0.2489 0.7511 55753 6267 2567 2052 0.1327
0.3174 8.0 11656 0.4269 0.1676 0.1624 0.2482 0.7518 55846 6299 2442 2083 0.1326
0.277 9.0 13113 0.4332 0.1700 0.1644 0.2505 0.7495 55831 6331 2425 2227 0.1346
0.2625 10.0 14570 0.4380 0.1700 0.1642 0.2501 0.7499 55894 6327 2366 2286 0.1345

Framework versions

  • Transformers 4.21.2
  • Pytorch 1.12.1+cu116
  • Datasets 2.4.0
  • Tokenizers 0.12.1