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: 0.1691
Mer: 0.1635
Wil: 0.2493
Wip: 0.7507
Hits: 55884
Substitutions: 6313
Deletions: 2390
Insertions: 2217
Cer: 0.1333

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: 10
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
0.5757	1.0	1457	0.4697	0.2091	0.1971	0.2856	0.7144	55012	6650	2925	3927	0.1749
0.543	2.0	2914	0.4228	0.1857	0.1777	0.2657	0.7343	55522	6547	2518	2932	0.1492
0.4448	3.0	4371	0.4108	0.1728	0.1670	0.2537	0.7463	55649	6380	2558	2220	0.1353
0.3968	4.0	5828	0.4075	0.1705	0.1649	0.2508	0.7492	55762	6314	2511	2187	0.1336
0.3529	5.0	7285	0.4146	0.1711	0.1651	0.2515	0.7485	55875	6370	2342	2337	0.1369
0.3135	6.0	8742	0.4188	0.1687	0.1633	0.2490	0.7510	55820	6291	2476	2128	0.1337
0.3405	7.0	10199	0.4271	0.1694	0.1637	0.2497	0.7503	55889	6324	2374	2242	0.1325
0.3131	8.0	11656	0.4301	0.1688	0.1634	0.2495	0.7505	55835	6330	2422	2152	0.1315
0.2796	9.0	13113	0.4342	0.1688	0.1633	0.2493	0.7507	55886	6327	2374	2203	0.1335
0.2722	10.0	14570	0.4380	0.1691	0.1635	0.2493	0.7507	55884	6313	2390	2217	0.1333

Framework versions

Transformers 4.21.2
Pytorch 1.12.1+cu116
Datasets 2.4.0
Tokenizers 0.12.1