MubarakB/whisper-base-en-v1

Whisper base Luganda

This model is a fine-tuned version of openai/whisper-base.en on the Sunbird dataset. It achieves the following results on the evaluation set:

• Loss: 0.6134
• Wer: 40.8010
• Cer: 10.6921

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: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3459	6.3694	1000	0.5885	45.1815	14.4972
0.0532	12.7389	2000	0.5441	38.6733	10.0723
0.0108	19.1083	3000	0.6118	39.9249	10.2789
0.0044	25.4777	4000	0.6134	40.8010	10.6921

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1