moatazlumin/arabic_corrector_addresses

arabic_auto_correction_byt5

This model is a fine-tuned version of google/byt5-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0016
• Cer: 0.01
• Bleu: 0.9381

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: 2e-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
• num_epochs: 16

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Bleu
No log	1.0	1000	0.0049	0.03	0.7215
1.1488	2.0	2000	0.0034	0.02	0.8014
1.1488	3.0	3000	0.0026	0.01	0.8477
0.0041	4.0	4000	0.0022	0.01	0.8757
0.0041	5.0	5000	0.0019	0.01	0.8919
0.0026	6.0	6000	0.0018	0.01	0.8988
0.0026	7.0	7000	0.0017	0.01	0.9118
0.0018	8.0	8000	0.0016	0.01	0.9173
0.0018	9.0	9000	0.0016	0.01	0.9219
0.0013	10.0	10000	0.0016	0.01	0.9249
0.0013	11.0	11000	0.0016	0.01	0.9284
0.001	12.0	12000	0.0017	0.01	0.9318
0.001	13.0	13000	0.0016	0.01	0.9354
0.0008	14.0	14000	0.0016	0.01	0.9344
0.0008	15.0	15000	0.0016	0.01	0.9386
0.0006	16.0	16000	0.0016	0.01	0.9381

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1