transcriber_sv-4.0

This model is a fine-tuned version of KBLab/kb-whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0051
• Wer: 0.32
• Accuracy: 97.8

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 200
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Accuracy
0.2385	0.5587	100	0.0891	2.66	83.33
0.0307	1.1173	200	0.0155	0.83	93.71
0.0045	1.6760	300	0.0050	0.36	98.11
0.0021	2.2346	400	0.0039	0.18	98.43
0.0007	2.7933	500	0.0040	0.22	98.43
0.0004	3.3520	600	0.0053	0.36	97.48
0.0028	3.9106	700	0.0047	0.36	97.48
0.0002	4.4693	800	0.0042	0.18	98.43
0.0012	5.0279	900	0.0051	0.32	97.8

Framework versions

• PEFT 0.18.1
• Transformers 5.2.0
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.2