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: 0.0062
• Wer: 0.0707
• Cer: 0.0287

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: 5e-06
• 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
0.0056	1.0	157	0.0061	0.0646	0.0262
0.0061	2.0	314	0.0060	0.0643	0.0263
0.0055	3.0	471	0.0059	0.0715	0.0321
0.0047	4.0	628	0.0060	0.0661	0.0272
0.0023	5.0	785	0.0061	0.0633	0.0257
0.0035	6.0	942	0.0063	0.0652	0.0267
0.0029	7.0	1099	0.0067	0.0659	0.0274
0.0017	8.0	1256	0.0070	0.0652	0.0270
0.0016	9.0	1413	0.0075	0.0668	0.0269
0.0009	10.0	1570	0.0079	0.0688	0.0282
0.0007	11.0	1727	0.0082	0.0708	0.0285
0.0003	12.0	1884	0.0085	0.0713	0.0375
0.0005	13.0	2041	0.0087	0.0720	0.0299
0.0002	14.0	2198	0.0089	0.0735	0.0381
0.0002	15.0	2355	0.0091	0.0742	0.0396
0.0003	16.0	2512	0.0092	0.0713	0.0299
0.0001	17.0	2669	0.0093	0.0711	0.0283
0.0003	18.0	2826	0.0093	0.0710	0.0281
0.0002	19.0	2983	0.0098	0.0987	0.0418
0.0001	19.8768	3120	0.0094	0.0708	0.0279

Framework versions

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