Qarib_arabic_keyword_extraction

This model is a fine-tuned version of qarib/bert-base-qarib60_860k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4027
• Precision: 0.5369
• Recall: 0.5937
• F1: 0.5638
• Accuracy: 0.9408

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.2196	1.0	750	0.1674	0.4656	0.4190	0.4411	0.9327
0.1374	2.0	1500	0.1559	0.4741	0.5255	0.4985	0.9366
0.0976	3.0	2250	0.1711	0.4901	0.5650	0.5249	0.9378
0.0676	4.0	3000	0.1928	0.4884	0.5557	0.5199	0.9363
0.0474	5.0	3750	0.2109	0.5313	0.5438	0.5375	0.9402
0.0342	6.0	4500	0.2414	0.5259	0.5754	0.5495	0.9389
0.024	7.0	5250	0.2527	0.5076	0.5881	0.5449	0.9382
0.0186	8.0	6000	0.3029	0.5379	0.5654	0.5513	0.9400
0.0143	9.0	6750	0.3154	0.5307	0.5862	0.5571	0.9398
0.0108	10.0	7500	0.3490	0.5491	0.5810	0.5646	0.9403
0.0078	11.0	8250	0.3550	0.5475	0.5929	0.5693	0.9412
0.0068	12.0	9000	0.3681	0.5360	0.6019	0.5670	0.9406
0.0049	13.0	9750	0.3873	0.5264	0.6048	0.5629	0.9402
0.004	14.0	10500	0.3987	0.5380	0.5937	0.5644	0.9407
0.0034	15.0	11250	0.4027	0.5369	0.5937	0.5638	0.9408

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0