README.md

metadata

language:
  - ko
license: apache-2.0
base_model: openai/whisper-small
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - GGarri/241113_newdata
metrics:
  - wer
model-index:
  - name: Whisper Small ko
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: customdata
          type: GGarri/241113_newdata
        metrics:
          - name: Wer
            type: wer
            value: 0.908879049172687

Whisper Small ko

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

• Loss: 0.0506
• Cer: 1.2584
• Wer: 0.9089

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: 32
• eval_batch_size: 8
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
1.1428	1.56	100	0.8829	14.7984	14.5304
0.3434	3.12	200	0.2469	2.0625	1.7828
0.0286	4.69	300	0.0447	1.6430	1.4099
0.011	6.25	400	0.0382	1.5148	1.1070
0.0067	7.81	500	0.0409	1.4915	1.0837
0.0042	9.38	600	0.0383	1.2118	0.9438
0.0018	10.94	700	0.0396	1.3866	1.0371
0.0007	12.5	800	0.0445	1.4682	1.0604
0.0004	14.06	900	0.0386	1.2584	0.9089
0.0002	15.62	1000	0.0431	1.1769	0.8273
0.0011	17.19	1100	0.0475	1.2701	0.9205
0.0019	18.75	1200	0.0453	1.4915	1.1419
0.0012	20.31	1300	0.0437	1.2701	0.9205
0.0013	21.88	1400	0.0454	1.3284	0.9205
0.0003	23.44	1500	0.0436	1.3400	0.9438
0.0001	25.0	1600	0.0460	1.3284	0.9904
0.0001	26.56	1700	0.0464	1.3517	1.0137
0.0001	28.12	1800	0.0464	1.3400	1.0021
0.0001	29.69	1900	0.0467	1.3167	0.9788
0.0001	31.25	2000	0.0468	1.3167	0.9788
0.0001	32.81	2100	0.0470	1.3284	0.9904
0.0001	34.38	2200	0.0473	1.2934	0.9438
0.0	35.94	2300	0.0475	1.3051	0.9555
0.0	37.5	2400	0.0477	1.3051	0.9555
0.0	39.06	2500	0.0478	1.3051	0.9555
0.0	40.62	2600	0.0480	1.2934	0.9438
0.0	42.19	2700	0.0482	1.2818	0.9322
0.0	43.75	2800	0.0483	1.2818	0.9322
0.0	45.31	2900	0.0485	1.2818	0.9322
0.0	46.88	3000	0.0486	1.2584	0.9089
0.0	48.44	3100	0.0487	1.2584	0.9089
0.0	50.0	3200	0.0489	1.2584	0.9089
0.0	51.56	3300	0.0490	1.2584	0.9089
0.0	53.12	3400	0.0491	1.2584	0.9089
0.0	54.69	3500	0.0492	1.2584	0.9089
0.0	56.25	3600	0.0493	1.2584	0.9089
0.0	57.81	3700	0.0493	1.2584	0.9089
0.0	59.38	3800	0.0495	1.2584	0.9089
0.0	60.94	3900	0.0495	1.2584	0.9089
0.0	62.5	4000	0.0496	1.2584	0.9089
0.0	64.06	4100	0.0499	1.2584	0.9089
0.0	65.62	4200	0.0501	1.2584	0.9089
0.0	67.19	4300	0.0502	1.2584	0.9089
0.0	68.75	4400	0.0504	1.2584	0.9089
0.0	70.31	4500	0.0505	1.2584	0.9089
0.0	71.88	4600	0.0506	1.2584	0.9089
0.0	73.44	4700	0.0506	1.2584	0.9089
0.0	75.0	4800	0.0506	1.2584	0.9089
0.0	76.56	4900	0.0506	1.2584	0.9089
0.0	78.12	5000	0.0506	1.2584	0.9089

Framework versions

• Transformers 4.39.2
• Pytorch 2.0.1
• Datasets 2.18.0
• Tokenizers 0.15.2