ahishamm/finetuned-whisper-quranic-large-v3

QURANIC Whisper Large V3

This model is a fine-tuned version of openai/whisper-large-v3 on the whisperQURANIC dataset. It achieves the following results on the evaluation set:

• Loss: 0.0238
• Wer: 268.8141

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1467	0.4	200	0.1302	42.9071
0.1226	0.8	400	0.0958	156.6683
0.0746	1.2	600	0.0772	494.4510
0.0868	1.6	800	0.0678	252.8552
0.0801	2.0	1000	0.0560	361.0673
0.0552	2.4	1200	0.0473	153.8658
0.053	2.8	1400	0.0399	310.5204
0.0421	3.2	1600	0.0308	305.3961
0.0291	3.6	1800	0.0266	242.5182
0.0303	4.0	2000	0.0238	268.8141

Framework versions

• Transformers 4.39.2
• Pytorch 2.2.0
• Datasets 2.18.0
• Tokenizers 0.15.1