whisper-large-sorani-v2

This model is a fine-tuned version of openai/whisper-large-v3 on the None 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: 3e-05
train_batch_size: 16
eval_batch_size: 8
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: 750
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.2256	0.3365	500	0.2478	42.8772
0.1959	0.6729	1000	0.2122	38.0271
0.1621	1.0094	1500	0.1883	34.3037
0.1345	1.3459	2000	0.1718	31.6828
0.1283	1.6824	2500	0.1607	29.1248
0.0689	2.0188	3000	0.1512	27.2824
0.0806	2.3553	3500	0.1542	27.9594
0.0827	2.6918	4000	0.1463	26.9971
0.0503	3.0283	4500	0.1472	25.0242
0.0417	3.3647	5000	0.1572	25.6141
0.0557	3.7012	5500	0.1466	24.4052
0.0294	4.0377	6000	0.1618	24.8017
0.0342	4.3742	6500	0.1564	23.3897
0.0349	4.7106	7000	0.1528	23.9942
0.0183	5.0471	7500	0.1598	24.6180
0.0264	5.3836	8000	0.1700	25.1402
0.0213	5.7201	8500	0.1660	22.6596
0.0113	6.0565	9000	0.1722	22.3211
0.0145	6.3930	9500	0.1791	23.3752
0.0146	6.7295	10000	0.1803	22.5048
0.0073	7.0659	10500	0.1766	21.4313
0.0081	7.4024	11000	0.1792	21.8617
0.0115	7.7389	11500	0.1792	21.6344
0.0051	8.0754	12000	0.1834	21.3153
0.0072	8.4118	12500	0.1829	21.3975
0.004	8.7483	13000	0.1846	21.1509
0.004	9.0848	13500	0.1845	21.1122
0.0053	9.4213	14000	0.1803	20.9768
0.004	9.7577	14500	0.1846	20.2176
0.002	10.0942	15000	0.1937	20.7930
0.0019	10.4307	15500	0.1940	20.4255
0.0029	10.7672	16000	0.1951	20.6190
0.0014	11.1036	16500	0.1897	19.8743
0.0009	11.4401	17000	0.1919	19.9662
0.0009	11.7766	17500	0.1990	19.7292
0.0013	12.1131	18000	0.1937	19.9275
0.0004	12.4495	18500	0.1974	19.2747
0.0006	12.7860	19000	0.2006	19.5116
0.0001	13.1225	19500	0.2016	18.9410
0.0002	13.4590	20000	0.2016	18.8346
0.0001	13.7954	20500	0.2016	18.4671
0.0002	14.1319	21000	0.2050	18.4091
0.0009	14.4684	21500	0.2074	18.4139
0.0	14.8048	22000	0.2079	18.3752