octava/whisper-small-indonesian-disaster

Whisper Small 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: 0.2947
• Wer: 11.3990

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: 32
• eval_batch_size: 32
• 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: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1756	1.6920	1000	0.2161	13.1418
0.0286	3.3841	2000	0.2258	11.6502
0.0065	5.0761	3000	0.2446	11.6659
0.0032	6.7682	4000	0.2531	11.2420
0.0022	8.4602	5000	0.2674	11.2577
0.001	10.1523	6000	0.2728	11.3990
0.0006	11.8443	7000	0.2819	11.7287
0.0004	13.5364	8000	0.2877	11.6502
0.0003	15.2284	9000	0.2927	11.5560
0.0003	16.9205	10000	0.2947	11.3990

Framework versions

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