./whisper-base-ea_8.5hr

This model is a fine-tuned version of openai/whisper-base on the Afrispeech-200 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7306
• Wer Ortho: 0.2818
• Wer: 0.2261
• Cer: 0.1004
• Precision: 0.8516
• Recall: 0.8552
• F1: 0.8527

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer	Cer	Precision	Recall	F1
1.0157	0.1825	100	0.9950	0.3122	0.2683	0.1220	0.8325	0.8364	0.8338
0.8347	0.3650	200	0.8313	0.3056	0.2562	0.1182	0.8429	0.8450	0.8433
0.6958	0.5474	300	0.7851	0.2905	0.2393	0.1060	0.8478	0.8503	0.8484
0.8021	0.7299	400	0.7532	0.2848	0.2286	0.1014	0.8541	0.8563	0.8546
0.8634	0.9124	500	0.7306	0.2818	0.2261	0.1004	0.8516	0.8552	0.8527

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1