bagasshw/whisper-tiny-javanese-openslr-v8

Whisper-Tiny-Java-v8

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2450
• Wer: 0.1763

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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_ratio: 0.1
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8116	1.6	1000	0.6226	0.4752
0.4683	3.2	2000	0.4166	0.4165
0.3047	4.8	3000	0.3430	0.3496
0.1679	6.4	4000	0.3180	0.2468
0.1405	8.0	5000	0.2874	0.2118
0.0856	9.6	6000	0.2819	0.2278
0.066	11.2	7000	0.2758	0.2034
0.0534	12.8	8000	0.2712	0.2234
0.044	14.4	9000	0.2690	0.2009
0.0393	16.0	10000	0.2669	0.1931
0.0344	17.6	11000	0.2601	0.1887
0.0292	19.2	12000	0.2627	0.1809
0.0272	20.8	13000	0.2597	0.1832
0.0231	22.4	14000	0.2556	0.1814
0.0223	24.0	15000	0.2562	0.1828
0.0192	25.6	16000	0.2534	0.1802
0.0167	27.2	17000	0.2512	0.1763
0.0178	28.8	18000	0.2496	0.1794
0.0143	30.4	19000	0.2461	0.1748
0.0147	32.0	20000	0.2450	0.1763

Framework versions

• Transformers 4.50.0.dev0
• Pytorch 2.7.0+cu128
• Datasets 2.16.0
• Tokenizers 0.21.1