transcriber_sv-3.0

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

• Loss: 0.0034
• Wer: 0.1
• Accuracy: 99.09

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
0.0985	0.5376	100	0.0274	1.56	89.09
0.0130	1.0753	200	0.0093	0.52	95.76
0.0080	1.6129	300	0.0055	0.24	97.88
0.0027	2.1505	400	0.0056	0.35	98.18
0.0024	2.6882	500	0.0044	0.17	98.48
0.0008	3.2258	600	0.0039	0.14	98.79
0.0006	3.7634	700	0.0035	0.14	98.79
0.0005	4.3011	800	0.0035	0.1	99.09
0.0004	4.8387	900	0.0034	0.1	99.09
0.0003	5.3763	1000	0.0034	0.1	99.09
0.0003	5.9140	1100	0.0034	0.1	99.09
0.0003	6.4516	1200	0.0034	0.1	99.09
0.0003	6.9892	1300	0.0034	0.1	99.09

Framework versions

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