whisper-small-quran-tashkeel

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

• Loss: 0.4243
• Wer: 41.3981
• Cer: 28.0299
• Wer Normalized: 39.8662

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: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 100
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Wer Normalized
0.0025	29.4179	1000	0.3436	37.7707	23.9281	36.3973
0.0002	58.8358	2000	0.3843	37.2777	23.1946	35.7281
0.0001	88.2388	3000	0.4053	39.4436	25.1808	37.8764
0.0001	117.6567	4000	0.4185	40.2007	26.173	38.6688
0.0	147.0597	5000	0.4243	41.3981	28.0299	39.8662

Framework versions

• Transformers 4.57.6
• Pytorch 2.10.0+cu128
• Datasets 4.6.1
• Tokenizers 0.22.2