PradyumSomebody/whisper-small-hi-custom5

Whisper Small EN - Pradyum Agarwal 4

This model is a fine-tuned version of openai/whisper-small on the Audio Medical Combined Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.0009
• Wer: 0.0

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: 10
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.749	0.5556	25	0.8143	17.8552
0.6169	1.1111	50	0.6652	14.0483
0.4324	1.6667	75	0.3086	12.4397
0.1351	2.2222	100	0.1857	8.6863
0.1414	2.7778	125	0.1251	5.3083
0.0638	3.3333	150	0.0827	2.7346
0.053	3.8889	175	0.0494	1.7694
0.0303	4.4444	200	0.0283	0.9115
0.0253	5.0	225	0.0155	0.5362
0.012	5.5556	250	0.0111	0.4826
0.0119	6.1111	275	0.0051	0.1072
0.0028	6.6667	300	0.0035	0.0
0.0066	7.2222	325	0.0030	0.0536
0.0045	7.7778	350	0.0026	0.0536
0.0015	8.3333	375	0.0020	0.0536
0.0013	8.8889	400	0.0018	0.0
0.0015	9.4444	425	0.0015	0.0
0.0016	10.0	450	0.0011	0.0
0.0008	10.5556	475	0.0010	0.0
0.0007	11.1111	500	0.0009	0.0

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1