README.md

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-kor3_de_all
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: whisper-kor3_de_all
          type: arrow
          config: default
          split: train
          args: 'config: ko, split: valid'
        metrics:
          - name: Wer
            type: wer
            value: 17.590945836701696

whisper-kor3_de_all

This model is a fine-tuned version of openai/whisper-small on the whisper-kor3_de_all dataset. It achieves the following results on the evaluation set:

• Loss: 0.2446
• Wer: 17.5909
• Cer: 7.8655

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
0.2987	0.05	100	0.2906	19.7898	9.3285
0.2658	0.09	200	0.2795	19.3371	9.6942
0.2748	0.14	300	0.2774	19.4341	8.9980
0.279	0.18	400	0.2767	22.5061	10.6901
0.2634	0.23	500	0.2837	19.7736	8.9319
0.2816	0.28	600	0.2826	19.8868	9.2315
0.2698	0.32	700	0.2826	19.8222	8.9759
0.2728	0.37	800	0.2794	19.9030	8.9187
0.2951	0.42	900	0.2752	20.1778	9.2271
0.2853	0.46	1000	0.2754	19.6281	9.3637
0.264	0.51	1100	0.2769	19.8222	9.1434
0.2684	0.55	1200	0.2745	19.8545	9.1390
0.286	0.6	1300	0.2731	19.6766	8.9627
0.2636	0.65	1400	0.2725	19.3048	8.7512
0.262	0.69	1500	0.2690	19.6281	8.9848
0.262	0.74	1600	0.2698	19.9515	9.1610
0.2788	0.78	1700	0.2693	19.7251	9.2491
0.2606	0.83	1800	0.2636	18.7065	8.6807
0.2601	0.88	1900	0.2626	18.9329	8.9231
0.249	0.92	2000	0.2649	19.0137	8.7777
0.2594	0.97	2100	0.2598	18.0922	8.1519
0.1764	1.02	2200	0.2565	17.8658	8.1123
0.1603	1.06	2300	0.2556	18.3508	8.2401
0.1572	1.11	2400	0.2561	19.1269	9.3549
0.1536	1.15	2500	0.2564	18.1568	8.1872
0.1719	1.2	2600	0.2543	18.0598	8.2665
0.1543	1.25	2700	0.2557	17.9143	8.1431
0.1636	1.29	2800	0.2519	17.8173	8.0991
0.1672	1.34	2900	0.2507	18.3670	8.6851
0.1519	1.39	3000	0.2528	18.8844	8.8834
0.1582	1.43	3100	0.2502	17.9143	8.1387
0.164	1.48	3200	0.2507	18.1083	8.3238
0.1464	1.52	3300	0.2487	18.1407	8.2973
0.1492	1.57	3400	0.2473	18.0760	8.2929
0.149	1.62	3500	0.2467	17.9143	8.1343
0.1592	1.66	3600	0.2457	17.9628	8.2753
0.1533	1.71	3700	0.2449	17.8173	7.9933
0.1597	1.75	3800	0.2454	17.8011	8.1475
0.1293	1.8	3900	0.2448	17.6233	7.8655
0.1499	1.85	4000	0.2446	17.5909	7.8655

Framework versions

• Transformers 4.33.2
• Pytorch 2.1.0+cu121
• Datasets 2.14.5
• Tokenizers 0.13.3