whisper-medium-Split-Sentences-filterEnglish

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

• Loss: 0.7103
• Cer: 16.1644
• Wer: 28.2682

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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: 1000
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
2.4948	1.0	1308	0.7037	25.9766	43.2308
0.9343	2.0	2616	0.6405	23.4593	39.0262
0.7284	3.0	3924	0.6153	23.8799	38.0744
0.6054	4.0	5232	0.6060	20.6629	35.0476
0.5082	5.0	6540	0.6058	20.7068	34.4587
0.4317	6.0	7848	0.6033	18.0619	30.3979
0.3693	7.0	9156	0.6054	18.8591	31.5620
0.3154	8.0	10464	0.6141	17.7709	30.4253
0.2705	9.0	11772	0.6353	18.1157	30.6375
0.2304	10.0	13080	0.6409	17.3304	30.1787
0.1978	11.0	14388	0.6458	17.0075	30.0007
0.1712	12.0	15696	0.6634	17.7768	30.4526
0.1459	13.0	17004	0.6766	17.1231	29.8500
0.127	14.0	18312	0.6921	16.6288	28.9393
0.1091	15.0	19620	0.6991	16.4753	28.7612
0.0958	16.0	20928	0.7083	16.3557	28.4120
0.0843	17.0	22236	0.7103	16.1644	28.2682
0.0742	18.0	23544	0.7253	16.2242	28.1586
0.0663	19.0	24852	0.7321	17.1151	29.1515
0.0601	20.0	26160	0.7392	16.5152	28.2545
0.0548	21.0	27468	0.7428	16.4115	28.4120
0.0503	22.0	28776	0.7543	16.2182	28.1791

Framework versions

• Transformers 4.53.3
• Pytorch 2.7.1+cu118
• Datasets 3.6.0
• Tokenizers 0.21.2