whispertiny_buah25sil_v1_2

This model is a fine-tuned version of openai/whisper-tiny on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.1199
Wer: 6.0708

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 150000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0148	0.0133	2000	0.0720	11.3670
0.0104	0.0267	4000	0.0463	7.4733
0.0075	0.04	6000	0.0406	6.5732
0.0064	0.0533	8000	0.0575	7.5361
0.0055	0.0667	10000	0.0578	7.5361
0.0051	1.0036	12000	0.0576	7.2012
0.0034	1.0169	14000	0.0638	7.1384
0.0036	1.0303	16000	0.0596	7.2221
0.0031	1.0436	18000	0.0591	6.4476
0.0027	1.0569	20000	0.0722	6.8453
0.0027	1.0703	22000	0.0654	6.8034
0.0022	2.0072	24000	0.0536	5.8824
0.0023	2.0205	26000	0.0845	7.1384
0.0017	2.0339	28000	0.0740	6.5941
0.0021	2.0472	30000	0.0696	6.4476
0.0017	2.0605	32000	0.0706	6.1336
0.0013	2.0739	34000	0.0886	7.0546
0.0017	3.0108	36000	0.0715	6.5104
0.0012	3.0242	38000	0.0767	6.8453
0.0016	3.0375	40000	0.0806	6.9500
0.0014	3.0508	42000	0.0930	7.7036
0.0014	3.0642	44000	0.0789	6.7406
0.0011	4.0011	46000	0.0838	6.7197
0.001	4.0144	48000	0.0754	6.1545
0.0013	4.0278	50000	0.0682	6.0917
0.0008	4.0411	52000	0.0866	6.6778
0.0008	4.0544	54000	0.0971	7.1593
0.0007	4.0678	56000	0.0754	5.9452
0.0006	5.0047	58000	0.0789	6.0917
0.0012	5.0180	60000	0.0853	6.4266
0.0009	5.0314	62000	0.0756	6.1126
0.0006	5.0447	64000	0.0792	6.0917
0.0007	5.0580	66000	0.0860	6.3638
0.0006	5.0714	68000	0.0998	6.6360
0.0006	6.0083	70000	0.0861	6.4894
0.0005	6.0216	72000	0.0657	5.3381
0.0004	6.0350	74000	0.0811	6.0080
0.0005	6.0483	76000	0.0861	5.8614
0.0005	6.0616	78000	0.0830	5.9033
0.0006	6.0750	80000	0.0874	6.1545
0.0006	7.0119	82000	0.0775	5.5474
0.0005	7.0252	84000	0.0930	5.9242
0.0005	7.0386	86000	0.0866	6.0080
0.0005	7.0519	88000	0.0786	5.6311
0.0002	7.0652	90000	0.0842	5.7568
0.0002	8.0022	92000	0.1018	6.8872
0.0004	8.0155	94000	0.0887	5.9242
0.0003	8.0289	96000	0.0884	6.0080
0.0002	8.0422	98000	0.0968	6.1964
0.0003	8.0555	100000	0.0915	5.6521
0.0002	8.0689	102000	0.1022	5.8196
0.0004	9.0058	104000	0.0968	5.6730
0.0001	9.0191	106000	0.0948	5.7149
0.0004	9.0325	108000	0.0943	5.5055
0.0001	9.0458	110000	0.0984	5.2962
0.0001	9.0591	112000	0.1105	6.0289
0.0001	9.0725	114000	0.1058	5.9452
0.0	10.0094	116000	0.1007	5.4427
0.0001	10.0227	118000	0.1001	5.5055
0.0001	10.0361	120000	0.0923	5.4009
0.0002	10.0494	122000	0.1140	5.7568
0.0001	10.0627	124000	0.1091	5.8614
0.0001	10.0761	126000	0.1143	5.7986
0.0	11.0130	128000	0.1184	6.0917
0.0	11.0263	130000	0.1186	6.1754
0.0	11.0397	132000	0.1094	5.7149
0.0	11.0530	134000	0.1193	5.9661
0.0	11.0663	136000	0.1182	5.9033
0.0	12.0033	138000	0.1224	6.2382
0.0	12.0166	140000	0.1216	6.0289
0.0	12.0299	142000	0.1179	5.9452
0.0	12.0433	144000	0.1160	5.9242
0.0	12.0566	146000	0.1188	5.9452
0.0	12.0699	148000	0.1201	6.0708
0.0	13.0069	150000	0.1199	6.0708

Framework versions

Transformers 4.41.2
Pytorch 2.3.0+cu121
Datasets 4.0.0
Tokenizers 0.19.1

albertuspekerti/whispertiny_buah25sil_v1_2

whispertiny_buah25sil_v1_2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions