whisper-tiny-kor_eng_tiny_oc_tx

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

• Loss: 1.5894
• Cer: 18.1943

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
0.5083	12.5	100	1.3425	16.8263
0.0123	25.0	200	1.3863	24.6238
0.0005	37.5	300	1.4230	25.3078
0.0002	50.0	400	1.4463	26.2654
0.0001	62.5	500	1.4674	26.4022
0.0001	75.0	600	1.4848	26.4022
0.0001	87.5	700	1.4955	21.3406
0.0	100.0	800	1.5067	19.5622
0.0	112.5	900	1.5133	19.0150
0.0	125.0	1000	1.5204	19.0150
0.0	137.5	1100	1.5262	20.2462
0.0	150.0	1200	1.5295	18.0575
0.0	162.5	1300	1.5329	18.0575
0.0	175.0	1400	1.5389	18.0575
0.0	187.5	1500	1.5404	17.6471
0.0	200.0	1600	1.5420	17.6471
0.0	212.5	1700	1.5450	18.3311
0.0	225.0	1800	1.5459	17.5103
0.0	237.5	1900	1.5473	17.6471
0.0	250.0	2000	1.5490	17.2367
0.0	262.5	2100	1.5582	17.2367
0.0	275.0	2200	1.5655	17.2367
0.0	287.5	2300	1.5687	17.2367
0.0	300.0	2400	1.5715	17.2367
0.0	312.5	2500	1.5745	17.2367
0.0	325.0	2600	1.5775	17.2367
0.0	337.5	2700	1.5796	17.2367
0.0	350.0	2800	1.5809	17.2367
0.0	362.5	2900	1.5833	17.6471
0.0	375.0	3000	1.5831	17.6471
0.0	387.5	3100	1.5851	17.6471
0.0	400.0	3200	1.5856	17.6471
0.0	412.5	3300	1.5857	17.6471
0.0	425.0	3400	1.5871	17.6471
0.0	437.5	3500	1.5885	18.1943
0.0	450.0	3600	1.5880	18.1943
0.0	462.5	3700	1.5883	18.1943
0.0	475.0	3800	1.5887	18.1943
0.0	487.5	3900	1.5897	18.1943
0.0	500.0	4000	1.5894	18.1943

Framework versions

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