metadata

language:
  - ko
license: apache-2.0
base_model: openai/whisper-small
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
datasets:
  - arrow
metrics:
  - wer
model-index:
  - name: whisper-kor_noising_full
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: whisper-kor_noising_full
          type: arrow
          config: default
          split: train
          args: 'config: ko, split: valid'
        metrics:
          - name: Wer
            type: wer
            value: 15.237651444547994

whisper-kor_noising_full

This model is a fine-tuned version of openai/whisper-small on the whisper-kor_noising_full dataset. It achieves the following results on the evaluation set:

Loss: 0.1966
Wer: 15.2377
Cer: 7.1689

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: 1e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1965	0.05	100	0.1848	13.3271	5.5988
0.2196	0.09	200	0.1880	13.7310	5.6454
0.2601	0.14	300	0.1932	14.8493	6.1532
0.2118	0.18	400	0.2005	15.8434	6.5002
0.2784	0.23	500	0.2088	16.0298	6.7160
0.2421	0.28	600	0.2105	16.0920	6.7160
0.2209	0.32	700	0.2159	16.9773	7.0884
0.2426	0.37	800	0.2157	17.2258	7.1096
0.2429	0.42	900	0.2166	16.7754	6.9403
0.258	0.46	1000	0.2158	17.2569	7.0673
0.2605	0.51	1100	0.2135	16.5113	6.9784
0.2196	0.55	1200	0.2120	16.7443	6.8261
0.2423	0.6	1300	0.2163	16.8841	7.0884
0.2389	0.65	1400	0.2138	16.6201	7.0419
0.2314	0.69	1500	0.2149	16.8531	6.8599
0.2509	0.74	1600	0.2126	17.2103	7.8206
0.2329	0.78	1700	0.2103	16.0764	6.7457
0.2504	0.83	1800	0.2092	15.8590	6.6526
0.2632	0.88	1900	0.2107	16.2783	6.8726
0.2374	0.92	2000	0.2091	16.3249	6.7245
0.2625	0.97	2100	0.2057	15.7658	6.5425
0.1471	1.02	2200	0.2052	15.8434	6.5129
0.1541	1.06	2300	0.2069	16.3249	6.7457
0.1301	1.11	2400	0.2042	15.9211	6.4917
0.1674	1.15	2500	0.2058	15.3153	6.4240
0.1435	1.2	2600	0.2060	15.6726	6.5044
0.1352	1.25	2700	0.2040	15.2998	6.3902
0.1258	1.29	2800	0.2019	15.1600	6.2971
0.1273	1.34	2900	0.2025	15.6881	6.4875
0.1527	1.39	3000	0.2031	15.7036	6.5044
0.1371	1.43	3100	0.2011	15.3308	6.3309
0.1247	1.48	3200	0.2003	15.2842	6.3521
0.1376	1.52	3300	0.1987	15.4551	7.2366
0.1194	1.57	3400	0.1999	15.5949	7.2704
0.144	1.62	3500	0.1983	14.9425	6.2886
0.1387	1.66	3600	0.1979	14.9425	6.2082
0.1372	1.71	3700	0.1979	15.3464	7.1435
0.1513	1.75	3800	0.1972	15.1445	7.0334
0.134	1.8	3900	0.1970	15.2377	7.1646
0.1165	1.85	4000	0.1966	15.2377	7.1689

Framework versions

Transformers 4.33.2
Pytorch 2.1.0+cu121
Datasets 2.14.5
Tokenizers 0.13.3