whisper-tiny-kor_eng_tiny_oc_ob

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4856
• Cer: 5.4945

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: 3e-05
• train_batch_size: 12
• eval_batch_size: 6
• seed: 42
• 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_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.5131	10.0	100	0.3638	8.4772
0.0156	20.0	200	0.4129	7.8493
0.0005	30.0	300	0.4288	6.9074
0.0002	40.0	400	0.4403	5.9655
0.0001	50.0	500	0.4499	5.9655
0.0001	60.0	600	0.4583	5.9655
0.0	70.0	700	0.4630	5.6515
0.0	80.0	800	0.4671	5.6515
0.0	90.0	900	0.4705	5.6515
0.0	100.0	1000	0.4732	5.6515
0.0	110.0	1100	0.4765	5.3375
0.0	120.0	1200	0.4780	5.3375
0.0	130.0	1300	0.4781	5.3375
0.0	140.0	1400	0.4796	5.3375
0.0	150.0	1500	0.4797	5.3375
0.0	160.0	1600	0.4802	5.3375
0.0	170.0	1700	0.4809	5.4945
0.0	180.0	1800	0.4807	5.4945
0.0	190.0	1900	0.4803	5.4945
0.0	200.0	2000	0.4797	5.4945
0.0	210.0	2100	0.4815	5.4945
0.0	220.0	2200	0.4820	5.4945
0.0	230.0	2300	0.4835	5.4945
0.0	240.0	2400	0.4839	5.4945
0.0	250.0	2500	0.4845	5.4945
0.0	260.0	2600	0.4849	5.4945
0.0	270.0	2700	0.4847	5.4945
0.0	280.0	2800	0.4852	5.4945
0.0	290.0	2900	0.4850	5.4945
0.0	300.0	3000	0.4861	5.4945
0.0	310.0	3100	0.4856	5.4945
0.0	320.0	3200	0.4857	5.4945
0.0	330.0	3300	0.4855	5.4945
0.0	340.0	3400	0.4854	5.4945
0.0	350.0	3500	0.4856	5.4945
0.0	360.0	3600	0.4851	5.4945
0.0	370.0	3700	0.4848	5.4945
0.0	380.0	3800	0.4859	5.4945
0.0	390.0	3900	0.4860	5.4945
0.0	400.0	4000	0.4856	5.4945

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.1