whisper-small-sesotho

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: 0.5607
• Wer: 0.4212

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: 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5799	8.4746	500	0.4938	0.4363
0.0106	16.9492	1000	0.5006	0.3685
0.0018	25.4237	1500	0.5144	0.3372
0.001	33.8983	2000	0.5209	0.4070
0.0002	42.3729	2500	0.5338	0.4846
0.0002	50.8475	3000	0.5435	0.4885
0.0001	59.3220	3500	0.5503	0.4632
0.0001	67.7966	4000	0.5554	0.4104
0.0001	76.2712	4500	0.5590	0.4217
0.0001	84.7458	5000	0.5607	0.4212

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2