mn720/swahili

Whisper Small Luganda

This model is a fine-tuned version of openai/whisper-small on the Common Voice 15.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3476
• Wer: 32.1768

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2853	0.11	500	0.4166	39.0969
0.2546	0.23	1000	0.4118	38.2528
0.2467	0.34	1500	0.3987	37.4636
0.2459	0.45	2000	0.3867	35.5323
0.2625	0.56	2500	0.3741	35.1086
0.2565	0.68	3000	0.3617	33.3147
0.2731	0.79	3500	0.3529	32.7463
0.2735	0.9	4000	0.3476	32.1768

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2