whisper-m-wo

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

• Loss: 0.4811
• Wer: 0.2595

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: 8
• 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: 40
• training_steps: 16000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7816	0.3912	1000	0.7274	0.6369
0.6368	0.7825	2000	0.6093	0.5042
0.3921	1.1737	3000	0.5506	0.4280
0.3494	1.5649	4000	0.5247	0.3115
0.3264	1.9562	5000	0.4907	0.3293
0.1734	2.3474	6000	0.4968	0.2973
0.1808	2.7387	7000	0.4811	0.2595
0.1064	3.1299	8000	0.4989	0.2490
0.0802	3.5211	9000	0.4975	0.2275
0.0745	3.9124	10000	0.4883	0.2429

Framework versions

• Transformers 4.43.3
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1