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: 0.0256
Wer: 0.1250
Cer: 0.0445

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-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: 20
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0263	0.9973	187	0.0205	0.1624	0.0613
0.0093	2.0	375	0.0149	0.1519	0.0529
0.0051	2.9973	562	0.0157	0.1580	0.0512
0.004	4.0	750	0.0181	0.1636	0.0539
0.002	4.9973	937	0.0193	0.1557	0.0502
0.0011	6.0	1125	0.0206	0.1558	0.0506
0.0009	6.9973	1312	0.0213	0.1513	0.0498
0.0005	8.0	1500	0.0214	0.1544	0.0504
0.0004	8.9973	1687	0.0220	0.1464	0.0458
0.0004	10.0	1875	0.0216	0.1459	0.0461
0.0002	10.9973	2062	0.0224	0.1452	0.0454
0.0001	12.0	2250	0.0224	0.1437	0.0452
0.0001	12.9973	2437	0.0234	0.2224	0.0832
0.0	14.0	2625	0.0231	0.1356	0.0540
0.0	14.9973	2812	0.0236	0.2134	0.0797
0.0	16.0	3000	0.0241	0.2159	0.0796
0.0	16.9973	3187	0.0253	0.1338	0.0517
0.0	18.0	3375	0.0257	0.1271	0.0493
0.0	18.9973	3562	0.0264	0.1287	0.0492
0.0	19.9467	3740	0.0266	0.1280	0.0489

Framework versions

Transformers 4.44.2
Pytorch 2.4.1+cu121
Datasets 3.2.0
Tokenizers 0.19.1