freshpearYoon/large-v3_3

whisper_finetune

This model is a fine-tuned version of openai/whisper-large-v3 on the aihub 한국어 아동 음성데이터 dataset. It achieves the following results on the evaluation set:

• Cer: 6.2655
• Loss: 1.0532
• Wer: 23.9347

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-08
• 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: 2001
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
1.5045	0.16	1000	6.8830	1.4103	26.6186
1.0745	0.32	2000	6.2655	1.0532	23.9347

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.17.0
• Tokenizers 0.15.2