gwanju2_mparameters1e-5__model

This model is a fine-tuned version of openai/whisper-medium on the Marcusxx/gwanju2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8405
• Cer: 19.4108

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• training_steps: 50000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.43	0.6046	1000	0.4009	166.0129
0.2587	1.2092	2000	0.3832	55.7046
0.2636	1.8138	3000	0.3843	98.7021
0.1623	2.4184	4000	0.4000	33.8374
0.1503	3.0230	5000	0.4133	27.1959
0.1061	3.6276	6000	0.4223	25.8623
0.0638	4.2322	7000	0.4534	24.9179
0.0671	4.8368	8000	0.4526	22.6665
0.0406	5.4414	9000	0.4991	21.5252
0.0372	6.0459	10000	0.5183	22.3878
0.0273	6.6505	11000	0.5348	21.8627
0.0195	7.2551	12000	0.5536	20.8458
0.0212	7.8597	13000	0.5466	21.0162
0.0146	8.4643	14000	0.5870	21.4089
0.0126	9.0689	15000	0.6040	26.2815
0.0113	9.6735	16000	0.6097	20.9229
0.0088	10.2781	17000	0.6190	21.1083
0.0098	10.8827	18000	0.6309	20.7387
0.0069	11.4873	19000	0.6489	20.5855
0.0056	12.0919	20000	0.6642	20.5590
0.0064	12.6965	21000	0.6514	20.8815
0.0053	13.3011	22000	0.6716	20.3149
0.005	13.9057	23000	0.6720	20.3068
0.0037	14.5103	24000	0.6912	20.6039
0.0044	15.1149	25000	0.7038	20.6166
0.004	15.7195	26000	0.6959	20.3149
0.0029	16.3241	27000	0.7064	20.6454
0.0031	16.9287	28000	0.7188	20.5463
0.002	17.5333	29000	0.7127	20.7640
0.0033	18.1378	30000	0.7193	20.4496
0.0019	18.7424	31000	0.7109	20.6719
0.002	19.3470	32000	0.7208	20.1306
0.0013	19.9516	33000	0.7442	20.1490
0.0007	20.5562	34000	0.7357	19.9198
0.0012	21.1608	35000	0.7501	19.5755
0.0022	21.7654	36000	0.7537	19.5870
0.0008	22.3700	37000	0.7686	19.6031
0.0013	22.9746	38000	0.7702	20.6097
0.0005	23.5792	39000	0.7712	19.9590
0.0007	24.1838	40000	0.7802	20.2020
0.0004	24.7884	41000	0.8050	19.5536
0.0004	25.3930	42000	0.8012	19.8093
0.0003	25.9976	43000	0.8049	19.6722
0.0002	26.6022	44000	0.8094	19.5214
0.0002	27.2068	45000	0.8109	19.6296
0.0001	27.8114	46000	0.8206	19.5248
0.0001	28.4160	47000	0.8299	19.5974
0.0001	29.0206	48000	0.8343	19.5490
0.0001	29.6252	49000	0.8395	19.4661
0.0001	30.2297	50000	0.8405	19.4108

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.2+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1