Whisper-WOLOF-10-hours-Google-Fleurs-dataset

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

• Loss: 1.4961
• Wer: 44.9192
• Cer: 16.7830

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: 32
• 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_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.1589	6.4935	500	1.0397	46.6260	18.8906
0.0602	12.9870	1000	1.2600	46.1173	17.4588
0.0042	19.4805	1500	1.3654	44.8401	16.6024
0.0013	25.9740	2000	1.4186	44.7383	16.5644
0.0007	32.4675	2500	1.4569	45.2922	17.2179
0.0006	38.9610	3000	1.4812	44.9531	16.6603
0.0005	45.4545	3500	1.4961	44.9192	16.7830

Framework versions

• Transformers 4.45.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.1