README.md

metadata

language:
  - ko
license: apache-2.0
base_model: INo0121/whisper-base-ko-callvoice
tags:
  - generated_from_trainer
datasets:
  - kresnik/zeroth_korean
metrics:
  - wer
model-index:
  - name: tmtms/whisper_checkpoints
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Zeroth-Korean
          type: kresnik/zeroth_korean
          args: 'config: ko, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 10.856798674898359

tmtms/whisper_checkpoints

This model is a fine-tuned version of INo0121/whisper-base-ko-callvoice on the Zeroth-Korean dataset. It achieves the following results on the evaluation set:

• Loss: 0.1501
• Wer: 10.8568

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: 24
• 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1644	1.08	1000	0.2571	22.2406
0.0822	2.16	2000	0.1818	14.2448
0.0528	3.23	3000	0.1575	11.1128
0.0383	4.31	4000	0.1501	10.8568

Framework versions

• Transformers 4.35.2
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.15.2