Mohamed2210/whisper-base-ar

Whisper base ar - Mohamed Ahmed-Mahmoud Nasser

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

• Loss: 0.1082
• Wer: 22.2325

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 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2541	0.5319	500	0.2250	44.5224
0.1526	1.0638	1000	0.1526	31.0242
0.1344	1.5957	1500	0.1321	27.8826
0.1217	2.1277	2000	0.1197	24.7066
0.1044	2.6596	2500	0.1153	23.7975
0.0886	3.1915	3000	0.1140	23.9471
0.1053	3.7234	3500	0.1090	22.3245
0.0843	4.2553	4000	0.1082	22.2325

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu121
• Datasets 3.3.2
• Tokenizers 0.21.0