OwLim/whisper-sundanese-java-finetune

Whisper Small Sundanese Java

This model is a fine-tuned version of OwLim/whisper-java-SLR41-SLR35 on the SLR44 Augmented Sundanese dataset. It achieves the following results on the evaluation set:

• Loss: 0.1298
• Wer: 11.3509

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
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5919	0.4292	100	0.2815	23.4139
0.395	0.8584	200	0.1782	15.8826
0.1788	1.2876	300	0.1554	13.0773
0.1654	1.7167	400	0.1445	11.4372
0.0581	2.1459	500	0.1337	11.5667
0.0572	2.5751	600	0.1335	11.6746
0.057	3.0043	700	0.1304	11.1135
0.0241	3.4335	800	0.1317	11.1135
0.0217	3.8627	900	0.1297	11.5019
0.0147	4.2918	1000	0.1298	11.3509

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1