Whisper Turbo Ko v0.0.1

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the 상담 음성 데이터셋 0.0.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1478
• Cer: 3.3711

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
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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	Cer
0.1318	1.9417	1000	0.1500	4.1320
0.0513	3.8835	2000	0.1380	4.3269
0.0215	5.8252	3000	0.1387	3.6299
0.0042	7.7670	4000	0.1478	3.3711

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1