nomnoos37/stt-turbo-1225-v1-full

Whisper Small ko

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

• Loss: 0.1905
• Wer: 12.1097

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: 4e-05
• train_batch_size: 64
• eval_batch_size: 64
• 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: 100
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.689	0.0107	10	1.0086	45.3169
0.0756	0.0214	20	0.6343	38.0322
0.0145	0.0322	30	0.6367	41.3434
0.0212	0.0429	40	0.7120	42.6679
0.0205	0.0536	50	0.4694	32.6395
0.016	0.0643	60	0.5533	38.7890
0.014	0.0750	70	0.4716	30.8420
0.0115	0.0857	80	0.6191	30.9366
0.0228	0.0965	90	0.7998	43.8978
0.0191	0.1072	100	0.7273	36.4238
0.026	0.1179	110	0.7720	42.3841
0.0196	0.1286	120	0.9171	79.4702
0.0178	0.1393	130	1.1460	136.0454
0.037	0.1501	140	0.5558	62.8193
0.0237	0.1608	150	0.6369	109.6500
0.0195	0.1715	160	0.6671	38.7890
0.0151	0.1822	170	0.6717	53.9262
0.0479	0.1929	180	0.5412	68.1173
0.0187	0.2036	190	0.5311	60.2649
0.0191	0.2144	200	0.4761	33.3964
0.0149	0.2251	210	0.6630	38.5998
0.0285	0.2358	220	0.6162	36.8023
0.0134	0.2465	230	0.5166	31.5043
0.0143	0.2572	240	0.6748	55.3453
0.0185	0.2680	250	0.5091	28.1930
0.0106	0.2787	260	0.4697	28.0984
0.0163	0.2894	270	0.4483	24.4087
0.0186	0.3001	280	0.3112	22.1381
0.018	0.3108	290	0.3752	26.7739
0.0067	0.3215	300	0.5734	28.0984
0.0129	0.3323	310	0.3768	22.3273
0.0196	0.3430	320	0.3069	23.4626
0.0096	0.3537	330	0.3197	20.5298
0.0143	0.3644	340	0.3839	43.8032
0.0082	0.3751	350	0.3098	80.1325
0.0099	0.3859	360	0.2946	77.6727
0.0146	0.3966	370	0.3007	19.3945
0.0115	0.4073	380	0.2685	17.3132
0.0058	0.4180	390	0.2686	16.7455
0.0067	0.4287	400	0.2572	15.6102
0.0095	0.4394	410	0.2400	14.9480
0.0085	0.4502	420	0.2436	15.2318
0.005	0.4609	430	0.2426	15.0426
0.0044	0.4716	440	0.2318	13.8127
0.0063	0.4823	450	0.2262	12.7720
0.0093	0.4930	460	0.2098	12.1097
0.0054	0.5038	470	0.2042	12.2990
0.0046	0.5145	480	0.1941	11.9205
0.0071	0.5252	490	0.1913	12.1097
0.0066	0.5359	500	0.1905	12.1097

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0