README.md

metadata

library_name: transformers
language:
  - ar
license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: Whisper tiny AR - BH
    results: []

Whisper tiny AR - BH

This model is a fine-tuned version of openai/whisper-tiny on the quran-ayat-speech-to-text dataset. It achieves the following results on the evaluation set:

• Loss: 0.0070
• Wer: 0.0780
• Cer: 0.0312

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: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• 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
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0099	1.0	157	0.0075	0.0746	0.0300
0.0068	2.0	314	0.0069	0.0682	0.0272
0.0054	3.0	471	0.0068	0.0700	0.0278
0.0043	4.0	628	0.0070	0.0726	0.0283
0.0031	5.0	785	0.0074	0.0726	0.0287
0.002	6.0	942	0.0082	0.0706	0.0285
0.0015	7.0	1099	0.0089	0.0758	0.0381
0.0005	8.0	1256	0.0096	0.0767	0.0479
0.0001	9.0	1413	0.0100	0.0767	0.0385
0.0002	10.0	1570	0.0104	0.0746	0.0474
0.0	11.0	1727	0.0107	0.0751	0.0474
0.0	12.0	1884	0.0109	0.0764	0.0482
0.0	13.0	2041	0.0112	0.0746	0.0479
0.0	14.0	2198	0.0113	0.0749	0.0488
0.0	15.0	2355	0.0116	0.0767	0.0495
0.0	16.0	2512	0.0119	0.0764	0.0488
0.0	17.0	2669	0.0121	0.0757	0.0487
0.0	18.0	2826	0.0122	0.0753	0.0489
0.0	19.0	2983	0.0127	0.0766	0.0390
0.0	19.8768	3120	0.0124	0.0755	0.0493

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0