Sirreajohn/whisper-tiny-minds14-en

Whisper Tiny MInDS-14

This model is a fine-tuned version of openai/whisper-tiny on the MInDS-14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6649
• Wer: 0.3276
• Wer Ortho: 0.3270

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: 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: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Wer Ortho
3.1843	0.8929	25	2.2625	0.4050	0.5262
1.3521	1.7857	50	0.5872	0.3849	0.4139
0.4052	2.6786	75	0.4956	0.3548	0.3646
0.2838	3.5714	100	0.4867	0.3306	0.3362
0.1607	4.4643	125	0.5027	0.3383	0.3492
0.0983	5.3571	150	0.5185	0.3264	0.3381
0.0545	6.25	175	0.5317	0.3235	0.3319
0.0285	7.1429	200	0.5650	0.3188	0.3300
0.0134	8.0357	225	0.5830	0.3217	0.3294
0.0064	8.9286	250	0.5854	0.3223	0.3251
0.0033	9.8214	275	0.6045	0.3259	0.3263
0.0022	10.7143	300	0.6145	0.3312	0.3319
0.0017	11.6071	325	0.6275	0.3182	0.3196
0.0013	12.5	350	0.6332	0.3247	0.3251
0.0011	13.3929	375	0.6397	0.3270	0.3276
0.0009	14.2857	400	0.6457	0.3247	0.3251
0.0008	15.1786	425	0.6506	0.3253	0.3263
0.0007	16.0714	450	0.6559	0.3253	0.3270
0.0007	16.9643	475	0.6604	0.3264	0.3270
0.0006	17.8571	500	0.6649	0.3276	0.3270

Framework versions

• Transformers 4.52.0
• Pytorch 2.8.0+cu126
• Datasets 3.6.0
• Tokenizers 0.21.4