simzacademy/whisper-small-tonga

Whisper small Tonga - Tonga ASR

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

• Loss: 0.5598
• Wer: 51.1127

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: 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
0.5631	0.9579	500	0.5425	56.3746
0.3865	1.9157	1000	0.4621	57.9972
0.243	2.8736	1500	0.4462	46.4070
0.1604	3.8314	2000	0.4600	47.2415
0.0831	4.7893	2500	0.4872	48.2151
0.0423	5.7471	3000	0.5048	56.5369
0.0195	6.7050	3500	0.5273	51.3213
0.0084	7.6628	4000	0.5476	48.9801
0.0051	8.6207	4500	0.5553	50.3245
0.0036	9.5785	5000	0.5598	51.1127

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2