whisper-ar-small-Version3

This model is a fine-tuned version of Moaaz5/whisper-ar-small-Version2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6456
• Wer: 31.0352

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: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 200
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8528	1.0	71	0.6243	38.6043
0.5861	2.0	142	0.5297	33.7708
0.4035	3.0	213	0.4987	31.9313
0.3142	4.0	284	0.5000	30.5397
0.1907	5.0	355	0.5212	30.5661
0.1353	6.0	426	0.5521	30.2340
0.1031	7.0	497	0.5959	30.8191
0.0525	8.0	568	0.6147	30.7453
0.042	9.0	639	0.6291	30.6241
0.0268	10.0	710	0.6456	31.0352

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0