whisper-tiny-minds14-finetuned

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: 0.6627
• Wer: 34.2381

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_FUSED 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: 50
• training_steps: 400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.9691	0.8929	25	2.3010	52.4368
1.2585	1.7857	50	0.7020	40.2838
0.4731	2.6786	75	0.5873	38.0012
0.3261	3.5714	100	0.5631	35.7804
0.2051	4.4643	125	0.5706	37.4460
0.134	5.3571	150	0.5751	35.2252
0.0852	6.25	175	0.5892	35.2869
0.0565	7.1429	200	0.6058	35.1635
0.0349	8.0357	225	0.6195	33.9297
0.0192	8.9286	250	0.6329	34.4849
0.0116	9.8214	275	0.6407	33.6829
0.0083	10.7143	300	0.6513	33.8063
0.0073	11.6071	325	0.6587	34.8550
0.0051	12.5	350	0.6596	34.2381
0.0044	13.3929	375	0.6617	34.1764
0.0043	14.2857	400	0.6627	34.2381

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu128
• Datasets 3.1.0
• Tokenizers 0.22.1