Mohamed2210/whisper-base-ar-upd

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: 0.1244
• Wer: 18.3084
• Cer: 8.3096

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: 16
• 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: cosine
• lr_scheduler_warmup_steps: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.107	1.0638	1000	0.1412	26.0759	10.2741
0.0927	2.1277	2000	0.1159	21.8412	9.1956
0.0601	3.1915	3000	0.1155	22.0368	9.2820
0.042	4.2553	4000	0.1135	18.7112	8.3240
0.018	5.3191	5000	0.1226	17.9517	8.1499
0.0068	6.3830	6000	0.1244	18.3084	8.3096

Framework versions

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