oyemade/whisper-small-yo

Whisper Small Yo - Oyemade Oyemaja

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16 dataset. It achieves the following results on the evaluation set:

• Loss: 1.1731
• Wer Ortho: 70.4834
• Wer: 49.8612

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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.2595	3.8462	500	0.7546	71.3700	52.0488
0.0312	7.6923	1000	0.9057	74.6210	53.1174
0.0134	11.5385	1500	1.0199	72.2090	51.7711
0.0059	15.3846	2000	1.0713	71.2842	51.6281
0.0087	19.2308	2500	1.1007	70.5787	50.1136
0.006	23.0769	3000	1.1568	70.8552	50.6100
0.0059	26.9231	3500	1.1327	69.0438	48.4645
0.0043	30.7692	4000	1.1731	70.4834	49.8612

Framework versions

• Transformers 4.40.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1