whisper-finetuned-v3_30e_augment_new

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

• Loss: 0.1077
• Wer: 52.3943
• Cer: 27.6678

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Use 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
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0913	1.0	1950	0.0906	64.8321	31.4175
0.0616	2.0	3900	0.0758	59.5771	29.2888
0.0428	3.0	5850	0.0732	57.7425	29.0100
0.0251	4.0	7800	0.0747	56.9652	28.8937
0.0227	5.0	9750	0.0780	56.0634	28.7220
0.0162	6.0	11700	0.0777	54.6331	28.5171
0.012	7.0	13650	0.0786	56.3122	28.6925
0.011	8.0	15600	0.0838	55.6592	28.4728
0.0069	9.0	17550	0.0810	55.0995	28.6703
0.0076	10.0	19500	0.0918	56.0323	28.5171
0.0048	11.0	21450	0.0918	54.4776	28.5060
0.0033	12.0	23400	0.0947	53.5759	28.2679
0.0035	13.0	25350	0.0876	54.7575	28.3805
0.0041	14.0	27300	0.0936	53.9801	28.1995
0.0023	15.0	29250	0.0943	52.8607	28.1146
0.0023	16.0	31200	0.0942	53.3271	28.2365
0.0025	17.0	33150	0.0986	53.2649	28.1829
0.0014	18.0	35100	0.0973	52.4565	28.0371
0.0008	19.0	37050	0.0970	53.0162	27.9189
0.0014	20.0	39000	0.1054	53.0784	27.9448
0.0009	21.0	40950	0.1016	52.4565	27.8192
0.001	22.0	42900	0.0991	52.7674	27.9928
0.0003	23.0	44850	0.1039	51.9590	27.7398
0.0003	24.0	46800	0.1071	52.8918	27.8968
0.0003	25.0	48750	0.1044	52.5498	27.7287
0.0001	26.0	50700	0.1085	52.0833	27.7897
0.0001	27.0	52650	0.1060	52.3632	27.8211
0.0001	28.0	54600	0.1082	52.7052	27.7306
0.0001	29.0	56550	0.1071	52.5187	27.8008
0.0	30.0	58500	0.1077	52.3943	27.6678

Framework versions

• Transformers 4.52.3
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1