whisper-swahili-large-v0.1

This model is a fine-tuned version of openai/whisper-large on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4162
• Wer: 26.3235

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
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1388	0.0681	250	0.6673	65.1304
0.4408	0.1362	500	0.5600	33.9259
0.3612	0.2042	750	0.4609	30.1521
0.3057	0.2723	1000	0.4162	26.3235

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.6.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.1