wh-ft-lr5e6-dtstf5-adm-ga1ba16-st15k-v2-evalstp500-pat5

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

• Loss: 0.5127
• Wer: 0.2195
• Cer: 0.1642

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: 16
• 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: 750
• training_steps: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5527	0.3047	500	0.5099	0.2741	0.2044
0.49	0.6094	1000	0.4720	0.2975	0.2388
0.4465	0.9141	1500	0.4545	0.2515	0.1923
0.3064	1.2188	2000	0.4499	0.2235	0.1698
0.3113	1.5235	2500	0.4480	0.2163	0.1616
0.3408	1.8282	3000	0.4396	0.2205	0.1591
0.1681	2.1328	3500	0.4677	0.2091	0.1542
0.1693	2.4375	4000	0.4602	0.2205	0.1604
0.1815	2.7422	4500	0.4660	0.2034	0.1520
0.091	3.0469	5000	0.5053	0.2092	0.1546
0.0949	3.3516	5500	0.5127	0.2195	0.1642

Framework versions

• Transformers 4.52.3
• Pytorch 2.7.0+cu118
• Datasets 3.5.1
• Tokenizers 0.21.1