ner-DarijaBERT-arabizi

This model is a fine-tuned version of SI2M-Lab/DarijaBERT-arabizi on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1011
• Precision: 0.7475
• Recall: 0.7753
• F1: 0.7611
• Accuracy: 0.9681

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: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	32	0.3240	0.5397	0.3251	0.4058	0.8996
No log	2.0	64	0.2541	0.5593	0.4720	0.5120	0.9203
No log	3.0	96	0.2062	0.5697	0.5828	0.5762	0.9350
No log	4.0	128	0.1791	0.6162	0.6313	0.6236	0.9426
No log	5.0	160	0.1528	0.6504	0.6803	0.6650	0.9509
No log	6.0	192	0.1308	0.6880	0.7262	0.7066	0.9582
No log	7.0	224	0.1189	0.7126	0.7270	0.7198	0.9612
No log	8.0	256	0.1100	0.7307	0.7661	0.7480	0.9651
No log	9.0	288	0.1037	0.7423	0.7567	0.7494	0.9667
No log	10.0	320	0.1011	0.7475	0.7753	0.7611	0.9681

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0