haseong8012/whisper-small-ko2

Whisper Small Ko - haseong8012

This model is a fine-tuned version of openai/whisper-small on the kresnik/zeroth_korean dataset. It achieves the following results on the evaluation set:

• Loss: 0.0981
• Wer: 6.7761
• Cer: 2.8934

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
0.0873	1.44	1000	0.1706	15.1333	6.0733
0.0274	2.87	2000	0.1186	9.6220	3.9887
0.0034	4.31	3000	0.1050	7.7699	3.1878
0.0011	5.75	4000	0.0990	7.4838	3.6079
0.0008	7.18	5000	0.0981	6.7761	2.8934

Framework versions

• Transformers 4.33.2
• Pytorch 1.12.1
• Datasets 2.14.5
• Tokenizers 0.13.3