arun100/whisper-small-ar-1

Whisper Small Arabic

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_16_0 ar dataset. It achieves the following results on the evaluation set:

• Loss: 0.4005
• Wer: 58.9073

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-06
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3404	1.53	500	0.4606	66.6216
0.2707	3.07	1000	0.4295	66.8500
0.2427	4.6	1500	0.4124	61.1662
0.2131	6.13	2000	0.4056	62.3038
0.2085	7.67	2500	0.4012	62.2754
0.1904	9.2	3000	0.3976	59.7341
0.1836	10.74	3500	0.4005	58.9073
0.1653	12.27	4000	0.3989	59.7774
0.1693	13.8	4500	0.3983	59.9462
0.1616	15.34	5000	0.3984	59.8300

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0