internalhell/whisper_small_ru_model_trainer_2

Whisper Small ru - slowlydoor

This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1989
• Wer: 16.6769
• Cer: 4.3640
• Ser: 59.1591

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: 8
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Ser
0.2176	0.1516	500	0.2575	21.0009	5.4512	69.0581
0.2146	0.3032	1000	0.2395	19.7826	5.2221	66.5785
0.1817	0.4548	1500	0.2264	18.5724	4.7800	64.4320
0.1862	0.6064	2000	0.2140	18.2088	4.7904	62.3542
0.1618	0.7580	2500	0.2049	17.0765	4.3953	60.4234
0.1597	0.9096	3000	0.1989	16.6769	4.3640	59.1591

Framework versions

• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1