numan98/whisper-tiny-nextayah-v5

Whisper Nextayah Tiny

This model is a fine-tuned version of tarteel-ai/whisper-tiny-ar-quran on the Synth verses dataset. It achieves the following results on the evaluation set:

• Loss: 0.0659
• Wer: 20.2519

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1019	1.0753	100	0.0978	26.5827
0.0481	2.1505	200	0.0843	24.7265
0.0137	3.2258	300	0.0715	21.5777
0.0042	4.3011	400	0.0659	20.2519

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0