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: 0.0206
- Wer: 0.1174
- Cer: 0.0425
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: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- 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 |
|---|---|---|---|---|---|
| 0.0204 | 1.0 | 250 | 0.0218 | 0.1251 | 0.0425 |
| 0.0116 | 2.0 | 500 | 0.0125 | 0.1298 | 0.0427 |
| 0.0086 | 3.0 | 750 | 0.0116 | 0.1229 | 0.0421 |
| 0.0049 | 4.0 | 1000 | 0.0121 | 0.1227 | 0.0449 |
| 0.0041 | 5.0 | 1250 | 0.0130 | 0.1231 | 0.0415 |
| 0.0029 | 6.0 | 1500 | 0.0143 | 0.1207 | 0.0407 |
| 0.0013 | 7.0 | 1750 | 0.0155 | 0.12 | 0.0390 |
| 0.0018 | 8.0 | 2000 | 0.0165 | 0.1265 | 0.0449 |
| 0.0008 | 9.0 | 2250 | 0.0173 | 0.1245 | 0.0414 |
| 0.0002 | 10.0 | 2500 | 0.0179 | 0.1222 | 0.0406 |
| 0.0002 | 11.0 | 2750 | 0.0182 | 0.1186 | 0.0400 |
| 0.0002 | 12.0 | 3000 | 0.0184 | 0.1198 | 0.0398 |
| 0.0001 | 13.0 | 3250 | 0.0187 | 0.1198 | 0.0404 |
| 0.0001 | 14.0 | 3500 | 0.0206 | 0.1174 | 0.0425 |
| 0.0 | 15.0 | 3750 | 0.0190 | 0.1189 | 0.0399 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.2.0
- Tokenizers 0.19.1