whisper-medium-bsbigcgen-male-model

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

• Loss: 0.6923
• Wer: 0.5177

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 500
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.039	0.6421	200	0.9764	0.6942
2.6473	1.2825	400	0.7705	0.5812
2.3913	1.9246	600	0.7010	0.5269
1.6283	2.5650	800	0.6923	0.5177
0.7783	3.2055	1000	0.7241	0.5131
0.8723	3.8475	1200	0.7207	0.5140
0.3633	4.4880	1400	0.7590	0.4961

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0