octava/whisper-small-indonesian-disaster-secondary

Whisper Medium Indonesian for Disaster Response

This model is a fine-tuned version of openai/whisper-small on the Indonesian Speech Dataset (InaVoCript, Fleurs, OpenSLR Javanese) dataset. It achieves the following results on the evaluation set:

• Loss: 1.1027
• Wer: 26.5584

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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
0.2835	1.3966	500	0.6987	26.6883
0.1078	2.7933	1000	0.7034	27.4675
0.023	4.1899	1500	0.8593	26.1039
0.0096	5.5866	2000	0.9625	26.9481
0.0051	6.9832	2500	0.9902	26.4935
0.0022	8.3799	3000	1.0349	25.7143
0.0016	9.7765	3500	1.0602	28.7338
0.0013	11.1732	4000	1.0808	26.2662
0.0012	12.5698	4500	1.0928	28.1494
0.0011	13.9665	5000	1.1027	26.5584

Framework versions

• Transformers 4.45.0
• Pytorch 2.8.0+cu129
• Datasets 3.0.1
• Tokenizers 0.20.0