whisper-medium-toigen-combined-model

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

• Loss: 0.6984
• Wer: 0.4649

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• 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
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.0083	0.8696	200	0.9494	0.6183
1.2927	1.7391	400	0.7467	0.5056
0.8214	2.6087	600	0.6984	0.4649
0.3719	3.4783	800	0.7200	0.4431
0.1984	4.3478	1000	0.7277	0.4285
0.0991	5.2174	1200	0.7420	0.4314

Framework versions

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