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: 0.0148
Wer: 0.1025
Cer: 0.0366

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: 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
0.0104	1.0	313	0.0092	0.1131	0.0389
0.009	2.0	626	0.0091	0.1104	0.0405
0.006	3.0	939	0.0093	0.1046	0.0371
0.0056	4.0	1252	0.0097	0.1061	0.0369
0.002	5.0	1565	0.0110	0.1061	0.0375
0.0021	6.0	1878	0.0110	0.1052	0.0464
0.0012	7.0	2191	0.0117	0.1019	0.0353
0.0008	8.0	2504	0.0126	0.1071	0.0457
0.0002	9.0	2817	0.0129	0.1019	0.0372
0.0002	10.0	3130	0.0140	0.1056	0.0380
0.0001	11.0	3443	0.0138	0.1003	0.0337
0.0001	12.0	3756	0.0142	0.0977	0.0327
0.0	13.0	4069	0.0146	0.1014	0.0346
0.0	14.0	4382	0.0152	0.0988	0.0340
0.0	15.0	4695	0.0164	0.1040	0.0390
0.0	16.0	5008	0.0161	0.1005	0.0336
0.0	17.0	5321	0.0165	0.1005	0.0330

Framework versions

Transformers 4.47.0
Pytorch 2.5.1+cu121
Datasets 3.2.0
Tokenizers 0.21.0