whisper-tiny-finetuned-minds14

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: 0.5953
• Wer Ortho: 0.3516
• Wer: 0.3371

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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 Ortho	Wer
3.9735	0.89	25	2.8501	0.5281	0.3979
1.8774	1.79	50	0.8237	0.4405	0.4067
0.53	2.68	75	0.5823	0.3874	0.3695
0.2962	3.57	100	0.5374	0.3726	0.3642
0.1982	4.46	125	0.5273	0.3658	0.3571
0.1361	5.36	150	0.5435	0.3701	0.3548
0.0711	6.25	175	0.5489	0.3609	0.3483
0.0387	7.14	200	0.5826	0.3664	0.3566
0.0221	8.04	225	0.5953	0.3516	0.3371
0.0123	8.93	250	0.6145	0.3510	0.3418
0.0061	9.82	275	0.6406	0.3597	0.3542
0.0041	10.71	300	0.6311	0.3479	0.3406
0.003	11.61	325	0.6513	0.3701	0.3619
0.0019	12.5	350	0.6630	0.3652	0.3613
0.0025	13.39	375	0.6672	0.3634	0.3601
0.0023	14.29	400	0.6738	0.3442	0.3418
0.0011	15.18	425	0.6746	0.3461	0.3436
0.0012	16.07	450	0.6788	0.3442	0.3430
0.0013	16.96	475	0.6865	0.3448	0.3447
0.0009	17.86	500	0.6921	0.3467	0.3459

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0