whisper-small-rw

This model is a fine-tuned version of mbazaNLP/Whisper-Small-Kinyarwanda on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0954
• Wer: 33.4204

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.4286	10	2.3900	34.2037
No log	2.8571	20	2.3111	34.5953
2.1422	4.2857	30	2.2492	34.2037
2.1422	5.7143	40	2.2022	33.9426
1.8742	7.1429	50	2.1669	33.8120
1.8742	8.5714	60	2.1406	33.6815
1.8742	10.0	70	2.1216	33.4204
1.7608	11.4286	80	2.1077	33.2898
1.7608	12.8571	90	2.0994	33.5509
1.6573	14.2857	100	2.0954	33.4204

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1