whisper-small-yoruba-07-17

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3346
• Wer Ortho: 34.5068
• Wer: 25.6765

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: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.7951	0.0745	250	0.8000	60.4651	45.9597
0.605	0.1490	500	0.6408	50.3176	39.3120
0.5273	0.2235	750	0.5621	45.8657	35.7526
0.4483	0.2980	1000	0.5138	43.4349	33.8902
0.4158	0.3725	1250	0.4752	41.9130	32.5295
0.4032	0.4470	1500	0.4434	41.1866	31.6249
0.3261	0.5215	1750	0.4153	40.2187	30.3193
0.3606	0.5959	2000	0.3910	38.0659	29.1049
0.3008	0.6704	2250	0.3769	36.7084	27.5409
0.2938	0.7449	2500	0.3608	36.2985	27.0924
0.2933	0.8194	2750	0.3494	35.6086	27.0448
0.277	0.8939	3000	0.3404	34.5474	25.5682
0.2849	0.9684	3250	0.3346	34.5068	25.6765

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1