Whisper-Base-Ar-Tashkeel-Dev - AzeemX

This model is a fine-tuned version of YazanSalameh/Whisper-base-Arabic on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0941
• Wer: 85.6472

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.1844	1.0323	1000	0.1771	99.3469
0.0986	2.0645	2000	0.1315	95.5247
0.0572	3.0968	3000	0.1108	87.4914
0.0355	4.1290	4000	0.0985	85.2242
0.0323	5.1613	5000	0.0941	85.6472

Framework versions

• Transformers 4.45.2
• Pytorch 2.3.1
• Datasets 3.0.1
• Tokenizers 0.20.0