bagasshw/whisper-tiny-javanese-openslr-v5

Whisper-Tiny-Java-v5

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: 0.2530
• 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: 50000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0859	0.4325	1000	0.9075	0.6419
0.7132	0.8651	2000	0.6100	0.5481
0.5258	1.2976	3000	0.4822	0.6352
0.4521	1.7301	4000	0.4058	0.4619
0.3848	1.0813	5000	0.4022	0.3778
0.351	1.2976	6000	0.3711	0.3326
0.3277	1.5138	7000	0.3546	0.3053
0.3122	1.7301	8000	0.3370	0.2862
0.3433	1.9464	9000	0.3173	0.2501
0.2336	2.1626	10000	0.3144	0.2563
0.2238	2.3789	11000	0.3043	0.2355
0.2225	2.5952	12000	0.2969	0.2403
0.218	2.8114	13000	0.2881	0.2326
0.1778	3.0277	14000	0.2848	0.2142
0.1669	3.2439	15000	0.2824	0.2114
0.1621	3.4602	16000	0.2812	0.2131
0.1585	3.6765	17000	0.2753	0.2114
0.1567	3.8927	18000	0.2723	0.1973
0.1092	4.1090	19000	0.2706	0.2005
0.1122	4.3253	20000	0.2704	0.2092
0.1138	4.5415	21000	0.2706	0.1959
0.121	4.7578	22000	0.2650	0.1952
0.11	4.9740	23000	0.2642	0.1935
0.0848	5.1903	24000	0.2655	0.1916
0.0844	5.4066	25000	0.2644	0.1890
0.0836	5.6228	26000	0.2626	0.1905
0.087	5.8391	27000	0.2587	0.1885
0.059	6.0554	28000	0.2594	0.1827
0.0596	6.2716	29000	0.2606	0.1835
0.0616	6.4879	30000	0.2587	0.1895
0.0634	6.7042	31000	0.2577	0.1805
0.0647	6.9204	32000	0.2557	0.1859
0.0467	7.1367	33000	0.2584	0.1800
0.0474	7.3529	34000	0.2545	0.1800
0.0478	7.5692	35000	0.2588	0.1827
0.0485	7.7855	36000	0.2559	0.1800
0.0456	8.0017	37000	0.2556	0.1804
0.0361	8.2180	38000	0.2560	0.1844
0.0354	8.4343	39000	0.2550	0.1806
0.0365	8.6505	40000	0.2557	0.1873
0.0388	8.8668	41000	0.2540	0.1843
0.0317	9.0830	42000	0.2547	0.1819
0.0334	9.2993	43000	0.2556	0.1780
0.033	9.5156	44000	0.2552	0.1801
0.0313	9.7318	45000	0.2540	0.1787
0.0318	9.9481	46000	0.2537	0.1772
0.0285	10.1644	47000	0.2534	0.1764
0.0256	10.3806	48000	0.2530	0.1771
0.0288	10.5969	49000	0.2532	0.1760
0.0265	10.8131	50000	0.2530	0.1763

Framework versions

• Transformers 4.50.0.dev0
• Pytorch 2.6.0+cu126
• Datasets 2.16.0
• Tokenizers 0.21.1