whisper-small-swahili

This model is a fine-tuned version of dmusingu/WHISPER-SMALL-SWAHILI-ASR-CV-14 on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.9641
• Model Preparation Time: 0.0073
• Wer: 26.3736

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: 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	Model Preparation Time	Wer
No log	1.4286	10	2.2013	0.0073	26.2515
No log	2.8571	20	2.1523	0.0073	26.3736
1.7887	4.2857	30	2.1129	0.0073	26.2515
1.7887	5.7143	40	2.0751	0.0073	26.2515
1.6873	7.1429	50	2.0428	0.0073	26.2515
1.6873	8.5714	60	2.0161	0.0073	26.3736
1.6873	10.0	70	1.9944	0.0073	26.3736
1.5626	11.4286	80	1.9788	0.0073	26.3736
1.5626	12.8571	90	1.9687	0.0073	26.3736
1.4991	14.2857	100	1.9641	0.0073	26.3736

Framework versions

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