whisper-tiny-minds14-enUS

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6872
• Wer: 0.2907
• Wer Ortho: 0.2907

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• 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
• lr_scheduler_warmup_steps: 200
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Wer Ortho
0.8922	1.7241	50	0.7444	0.3378	0.3378
0.3603	3.4483	100	0.5051	0.3048	0.3048
0.1591	5.1724	150	0.5103	0.3028	0.3028
0.0449	6.8966	200	0.5746	0.2941	0.2941
0.0104	8.6207	250	0.6113	0.3022	0.3022
0.0035	10.3448	300	0.6488	0.2867	0.2867
0.0011	12.0690	350	0.6746	0.2907	0.2907
0.0008	13.7931	400	0.6828	0.2894	0.2894
0.0007	15.5172	450	0.6872	0.2907	0.2907

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1