whisper-tiny-kor_eng_tiny_ed_is

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

• Loss: 3.6555
• Cer: 46.3895

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
0.7622	50.0	100	3.0720	41.2108
0.0056	100.0	200	3.2889	86.7980
0.0006	150.0	300	3.3519	85.7039
0.0003	200.0	400	3.3979	55.4340
0.0001	250.0	500	3.4321	55.0693
0.0001	300.0	600	3.4605	55.0693
0.0001	350.0	700	3.4788	55.0693
0.0001	400.0	800	3.4941	55.0693
0.0	450.0	900	3.5049	55.0693
0.0	500.0	1000	3.5161	55.0693
0.0	550.0	1100	3.5252	55.0693
0.0	600.0	1200	3.5331	55.0693
0.0	650.0	1300	3.5418	55.0693
0.0	700.0	1400	3.5452	54.5587
0.0	750.0	1500	3.5510	54.5587
0.0	800.0	1600	3.5575	55.6528
0.0	850.0	1700	3.5604	55.6528
0.0	900.0	1800	3.5692	55.6528
0.0	950.0	1900	3.5719	55.6528
0.0	1000.0	2000	3.5777	46.0248
0.0	1050.0	2100	3.5872	46.9730
0.0	1100.0	2200	3.5975	46.9730
0.0	1150.0	2300	3.6034	46.9730
0.0	1200.0	2400	3.6098	46.4624
0.0	1250.0	2500	3.6168	46.4624
0.0	1300.0	2600	3.6199	46.4624
0.0	1350.0	2700	3.6247	45.4413
0.0	1400.0	2800	3.6273	46.0248
0.0	1450.0	2900	3.6310	46.0248
0.0	1500.0	3000	3.6349	46.5354
0.0	1550.0	3100	3.6367	46.6083
0.0	1600.0	3200	3.6436	46.6083
0.0	1650.0	3300	3.6451	46.6083
0.0	1700.0	3400	3.6486	46.8271
0.0	1750.0	3500	3.6500	46.3895
0.0	1800.0	3600	3.6529	46.3895
0.0	1850.0	3700	3.6532	46.3895
0.0	1900.0	3800	3.6544	46.3895
0.0	1950.0	3900	3.6541	46.3895
0.0	2000.0	4000	3.6555	46.3895

Framework versions

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