bagasshw/whisper-tiny-javanese-openslr-v7

Whisper-Tiny-Java-v7

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: 1.6090
• Wer: 0.5104

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
0.9973	8.0	1000	1.0428	0.6821
0.3139	16.0	2000	0.8733	0.6234
0.0988	24.0	3000	0.8925	0.6347
0.0499	32.0	4000	0.9389	0.6149
0.0349	40.0	5000	0.9879	0.6206
0.0246	48.0	6000	1.0245	0.4681
0.0164	56.0	7000	1.1032	0.4715
0.0118	64.0	8000	1.1389	0.5985
0.0094	72.0	9000	1.1457	0.5793
0.0086	80.0	10000	1.1854	0.5375
0.0067	88.0	11000	1.2372	0.4664
0.005	96.0	12000	1.2564	0.4992
0.0043	104.0	13000	1.2725	0.4478
0.0039	112.0	14000	1.3331	0.5308
0.0036	120.0	15000	1.3492	0.6036
0.0028	128.0	16000	1.3910	0.4551
0.0034	136.0	17000	1.4366	0.4907
0.0024	144.0	18000	1.3797	0.6126
0.0019	152.0	19000	1.4368	0.5229
0.0017	160.0	20000	1.4499	0.5980
0.0014	168.0	21000	1.4369	0.5940
0.0015	176.0	22000	1.4615	0.5308
0.001	184.0	23000	1.4453	0.4986
0.0009	192.0	24000	1.4906	0.5471
0.0007	200.0	25000	1.4574	0.4941
0.0011	208.0	26000	1.4995	0.4647
0.0007	216.0	27000	1.5195	0.5155
0.0011	224.0	28000	1.4928	0.5483
0.0011	232.0	29000	1.5243	0.5144
0.0007	240.0	30000	1.5805	0.4884
0.0005	248.0	31000	1.5294	0.5116
0.0005	256.0	32000	1.5940	0.4975
0.0003	264.0	33000	1.5760	0.5003
0.0004	272.0	34000	1.5940	0.4873
0.0003	280.0	35000	1.6010	0.4681
0.0004	288.0	36000	1.5837	0.4845
0.0006	296.0	37000	1.5839	0.4794
0.0002	304.0	38000	1.5652	0.4754
0.0003	312.0	39000	1.6083	0.4833
0.0002	320.0	40000	1.5750	0.5189
0.0004	328.0	41000	1.6199	0.5980
0.0001	336.0	42000	1.5783	0.5353
0.0001	344.0	43000	1.5898	0.5099
0.0005	352.0	44000	1.6005	0.5833
0.0002	360.0	45000	1.5903	0.4873
0.0002	368.0	46000	1.6196	0.5150
0.0001	376.0	47000	1.6212	0.5251
0.0002	384.0	48000	1.6180	0.5539
0.0001	392.0	49000	1.6104	0.4963
0.0001	400.0	50000	1.6090	0.5104

Framework versions

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