bagasshw/whisper-tiny-javanese-openslr-v6

Whisper-Tiny-Java-v6

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: 1.0455
• Wer: 0.3609

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
1.1166	3.1952	1000	1.0200	0.6975
0.5644	6.3904	2000	0.7342	0.5898
0.2837	9.5856	3000	0.6663	0.5701
0.1451	12.7808	4000	0.6692	0.6157
0.0949	15.9760	5000	0.6929	0.6141
0.0621	19.1696	6000	0.7082	0.4790
0.0466	22.3648	7000	0.7456	0.4469
0.0338	25.56	8000	0.7601	0.4366
0.0288	28.7552	9000	0.7782	0.3894
0.0232	31.9504	10000	0.7977	0.4107
0.0212	35.144	11000	0.7976	0.4144
0.0178	38.3392	12000	0.8184	0.4011
0.0132	41.5344	13000	0.8311	0.3763
0.0145	44.7296	14000	0.8474	0.3790
0.0117	47.9248	15000	0.8625	0.4155
0.0101	51.1184	16000	0.8907	0.3758
0.0091	54.3136	17000	0.8973	0.3999
0.0087	57.5088	18000	0.9277	0.4183
0.0068	60.704	19000	0.9449	0.4389
0.0073	63.8992	20000	0.9372	0.3834
0.0066	67.0928	21000	0.9512	0.4038
0.0056	70.288	22000	0.9799	0.4063
0.0049	73.4832	23000	0.9893	0.3845
0.0046	76.6784	24000	0.9897	0.3809
0.0044	79.8736	25000	0.9970	0.3749
0.0039	83.0672	26000	1.0113	0.3761
0.0035	86.2624	27000	1.0149	0.3832
0.003	89.4576	28000	1.0032	0.3859
0.0025	92.6528	29000	1.0094	0.3857
0.0034	95.848	30000	1.0202	0.3733
0.0027	99.0416	31000	1.0113	0.3655
0.0023	102.2368	32000	1.0178	0.3767
0.002	105.432	33000	1.0110	0.3671
0.002	108.6272	34000	1.0301	0.3733
0.0018	111.8224	35000	1.0479	0.3850
0.002	115.016	36000	1.0216	0.3625
0.002	118.2112	37000	1.0406	0.3687
0.0011	121.4064	38000	1.0520	0.4130
0.0014	124.6016	39000	1.0564	0.3662
0.0009	127.7968	40000	1.0525	0.3788
0.0009	130.992	41000	1.0391	0.3598
0.0011	134.1856	42000	1.0483	0.3763
0.001	137.3808	43000	1.0376	0.4169
0.0008	140.576	44000	1.0424	0.3632
0.0007	143.7712	45000	1.0495	0.4160
0.0009	146.9664	46000	1.0547	0.3687
0.0005	150.16	47000	1.0583	0.3690
0.0007	153.3552	48000	1.0463	0.4123
0.0007	156.5504	49000	1.0464	0.3600
0.0005	159.7456	50000	1.0455	0.3609

Framework versions

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