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: 0.0081
  • Wer: 0.0885
  • Cer: 0.0350

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: 2
  • total_train_batch_size: 32
  • 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: 15
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
0.007 1.0 469 0.0069 0.0927 0.0352
0.0076 2.0 938 0.0065 0.0919 0.0353
0.0036 3.0 1407 0.0065 0.0905 0.0350
0.0041 4.0 1876 0.0068 0.0887 0.0340
0.003 5.0 2345 0.0071 0.0876 0.0330
0.003 6.0 2814 0.0076 0.0900 0.0358
0.0024 7.0 3283 0.0080 0.0916 0.0349
0.002 8.0 3752 0.0086 0.0889 0.0329
0.0015 9.0 4221 0.0089 0.1674 0.0708
0.0003 10.0 4690 0.0094 0.1690 0.0731
0.0005 11.0 5159 0.0097 0.1658 0.0705
0.0009 12.0 5628 0.0099 0.1669 0.0714
0.0005 13.0 6097 0.0101 0.1672 0.0712
0.0004 14.0 6566 0.0118 0.1017 0.0411
0.0001 15.0 7035 0.0103 0.1660 0.0713

Framework versions

  • Transformers 4.47.0
  • Pytorch 2.5.1+cu121
  • Datasets 3.2.0
  • Tokenizers 0.21.0
Downloads last month
83
Safetensors
Model size
37.8M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.

Model tree for Baselhany/Graduation_Project_Whisper_tiny

Finetuned
(1302)
this model