whisper-tiny-kor_eng_tiny_oc_ev

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

• Loss: 0.5052
• Cer: 8.7336

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
0.3603	5.5556	100	0.4065	11.2991
0.0352	11.1111	200	0.4413	9.4432
0.0014	16.6667	300	0.4427	9.2249
0.0004	22.2222	400	0.4449	8.4607
0.0002	27.7778	500	0.4504	8.4061
0.0001	33.3333	600	0.4548	8.4061
0.0001	38.8889	700	0.4587	7.9148
0.0001	44.4444	800	0.4621	7.9148
0.0001	50.0	900	0.4654	7.9148
0.0	55.5556	1000	0.4684	7.9148
0.0	61.1111	1100	0.4712	7.9148
0.0	66.6667	1200	0.4736	8.4607
0.0	72.2222	1300	0.4762	8.4061
0.0	77.7778	1400	0.4784	8.4061
0.0	83.3333	1500	0.4806	8.4607
0.0	88.8889	1600	0.4825	8.3515
0.0	94.4444	1700	0.4844	8.5699
0.0	100.0	1800	0.4859	8.5699
0.0	105.5556	1900	0.4878	8.5699
0.0	111.1111	2000	0.4893	8.5699
0.0	116.6667	2100	0.4909	8.5699
0.0	122.2222	2200	0.4921	8.6245
0.0	127.7778	2300	0.4938	8.7336
0.0	133.3333	2400	0.4949	8.7336
0.0	138.8889	2500	0.4955	8.7336
0.0	144.4444	2600	0.4968	8.7336
0.0	150.0	2700	0.4979	8.7336
0.0	155.5556	2800	0.4989	8.7336
0.0	161.1111	2900	0.5000	8.7882
0.0	166.6667	3000	0.5009	8.7882
0.0	172.2222	3100	0.5018	8.6245
0.0	177.7778	3200	0.5026	8.6245
0.0	183.3333	3300	0.5033	8.6245
0.0	188.8889	3400	0.5034	8.6245
0.0	194.4444	3500	0.5036	8.6245
0.0	200.0	3600	0.5041	8.6245
0.0	205.5556	3700	0.5046	8.7336
0.0	211.1111	3800	0.5051	8.7336
0.0	216.6667	3900	0.5049	8.7336
0.0	222.2222	4000	0.5052	8.7336

Framework versions

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