README.md

metadata

base_model: openai/whisper-tiny
datasets:
  - mozilla-foundation/common_voice_17_0
language:
  - ko
license: apache-2.0
metrics:
  - wer
tags:
  - generated_from_trainer
model-index:
  - name: Whisper Tiny Ko - Roooy
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: Common Voice 17.0
          type: mozilla-foundation/common_voice_17_0
          config: ko
          split: None
          args: 'config: ko, split: train+valid & test'
        metrics:
          - type: wer
            value: 61.8411000763942
            name: Wer

Whisper Tiny Ko - Roooy

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.3774
• Cer: 30.0833
• Wer: 61.8411
• Cer Wer Avg: 45.9622

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: 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Cer Wer Avg
0.0005	22.2222	1000	1.0691	27.3648	58.8235	43.0942
0.0002	44.4444	2000	1.1396	30.6350	62.9488	46.7919
0.0001	66.6667	3000	1.1884	31.1967	63.5982	47.3974
0.0001	88.8889	4000	1.2300	31.4776	64.2093	47.8435
0.0	111.1111	5000	1.2656	31.7284	64.7441	48.2362
0.0	133.3333	6000	1.2993	32.1396	65.0497	48.5946
0.0	155.5556	7000	1.3272	32.3804	64.9351	48.6577
0.0	177.7778	8000	1.3518	29.9829	61.8029	45.8929
0.0	200.0	9000	1.3693	30.1836	61.8793	46.0314
0.0	222.2222	10000	1.3774	30.0833	61.8411	45.9622

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1