whisper_kr_custom_split

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

Model description

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 5e-06
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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: cosine
lr_scheduler_warmup_steps: 1000
num_epochs: 1000

Training Loss	Epoch	Step	Validation Loss	Cer
0.0013	27.0015	1000	0.2591	0.1366
0.0001	54.0029	2000	0.2833	0.1782
0.0001	81.0044	3000	0.2992	0.1836
0.0	108.0058	4000	0.3125	0.1823
0.0	135.0073	5000	0.3257	0.1802
0.0	162.0087	6000	0.3387	0.1793
0.0	189.0102	7000	0.3525	0.1725
0.0	216.0116	8000	0.3671	0.1663
0.0003	243.0131	9000	0.2761	0.1624
0.0	270.0145	10000	0.2977	0.1686
0.0	297.016	11000	0.3107	0.1650
0.0	324.0175	12000	0.3214	0.1622
0.0	351.0189	13000	0.3306	0.1606
0.0	378.0204	14000	0.3391	0.1589
0.0	405.0218	15000	0.3476	0.1608
0.0	432.0233	16000	0.3561	0.1622
0.0	459.0247	17000	0.3636	0.1632
0.0	486.0262	18000	0.3728	0.1634
0.0	513.0276	19000	0.3819	0.1554
0.0	540.0291	20000	0.3916	0.1453
0.0	567.0305	21000	0.4012	0.1348
0.0	594.032	22000	0.4098	0.1270
0.0	621.0335	23000	0.4168	0.1244
0.0	648.0349	24000	0.4233	0.1233
0.0	675.0364	25000	0.4303	0.1214
0.0	702.0378	26000	0.4346	0.1227
0.0	729.0393	27000	0.4410	0.1244
0.0	756.0407	28000	0.4482	0.1251
0.0	783.0422	29000	0.4552	0.1249
0.0	810.0436	30000	0.4631	0.1262
0.0	837.0451	31000	0.4720	0.1298
0.0	864.0465	32000	0.4814	0.1279
0.0	891.048	33000	0.4888	0.1270
0.0	918.0495	34000	0.4973	0.1288
0.0	945.0509	35000	0.5055	0.1305
0.0	972.0524	36000	0.5135	0.1316
0.0	999.0538	37000	0.5203	0.1333