whisper-tiny-finetune-Moroccan-darija/README.md

library_name, language, license, base_model, tags, datasets, metrics, model-index

library_name	language	license	base_model	tags	datasets	metrics	model-index
transformers	ar	apache-2.0	openai/whisper-tiny	generated_from_trainer	atlasia/darija_bible_aligned	wer	name results whisper-tiny-finetune-Moroccan-darija-Anas task dataset metrics name type Automatic Speech Recognition automatic-speech-recognition name type darija_bible_aligned atlasia/darija_bible_aligned name type value Wer wer 5.2240026160889474

whisper-tiny-finetune-Moroccan-darija-Anas

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

• Loss: 0.0494
• Wer: 5.2240

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: 10
• 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: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1766	0.7037	1000	0.1846	20.2502
0.0822	1.4075	2000	0.1040	11.5558
0.0402	2.1112	3000	0.0713	7.8442
0.0331	2.8149	4000	0.0556	5.8004
0.0201	3.5186	5000	0.0494	5.2240

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1