This model is a fine-tuned version of openai/whisper-tiny on the haseong8012/korean-child-command-voice_train-0-10000 dataset. It achieves the following results on the evaluation set:

Loss: 0.1263
Wer: 27.1876
Cer: 19.8556

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: 5e-05
train_batch_size: 32
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: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.1915	2.0	500	0.3624	78.8440	158.9770
0.0459	4.0	1000	0.2271	108.7274	129.0652
0.0111	6.0	1500	0.1616	44.8481	48.1227
0.0035	8.0	2000	0.1742	46.9043	44.5694
0.0014	10.0	2500	0.1635	32.8536	33.4027
0.0003	12.0	3000	0.1325	25.9995	21.7572
0.0001	14.0	3500	0.1300	27.1190	25.5743
0.0001	16.0	4000	0.1285	27.3475	24.4084
0.0001	18.0	4500	0.1266	27.1190	19.7030
0.0001	20.0	5000	0.1263	27.1876	19.8556

Framework versions

Transformers 4.33.2
Pytorch 1.12.1
Datasets 2.14.5
Tokenizers 0.13.3

haseong8012/whisper-tiny_child-10k-sr48k

openai/whisper-tiny-Ko-haseong8012/korean-child-command-voice_train-0-10000

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

haseong8012/
whisper-tiny_child-10k-sr48k