bert-base-arabertv2_D3Lex_CE_19levels

This model is a fine-tuned version of aubmindlab/bert-base-arabertv2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8342
• Macro F1: 0.4364
• Macro Precision: 0.4424
• Macro Recall: 0.4454
• Accuracy: 0.5140
• Accuracy With Margin: 0.6744
• Distance: 1.2527
• Quadratic weighted kappa: 0.7889
• Accuracy 7: 0.6179
• Accuracy 5: 0.6653
• Accuracy 3: 0.7363

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: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Macro Precision	Macro Recall	Accuracy	Accuracy With Margin	Distance	Quadratic weighted kappa	Accuracy 7	Accuracy 5	Accuracy 3
1.8254	1.0	857	1.4573	0.3478	0.3741	0.3804	0.5044	0.6513	1.3171	0.7790	0.6223	0.6773	0.7415
1.2699	2.0	1714	1.4409	0.4133	0.4330	0.4316	0.5157	0.6688	1.2666	0.7825	0.6230	0.6802	0.7497
1.0213	3.0	2571	1.4714	0.4157	0.4210	0.4276	0.5272	0.6711	1.2473	0.7889	0.6308	0.6780	0.7458
0.8308	4.0	3428	1.6169	0.4074	0.4422	0.4110	0.5171	0.6724	1.2573	0.7848	0.6194	0.6663	0.7397
0.6128	5.0	4285	1.7657	0.4222	0.4375	0.4349	0.5164	0.6737	1.2575	0.7861	0.6208	0.6703	0.7391
0.4786	6.0	5142	1.8342	0.4364	0.4424	0.4454	0.5140	0.6744	1.2527	0.7889	0.6179	0.6653	0.7363

Framework versions

• Transformers 4.53.2
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.2