AraModernBert-coherence-clean

This model is a fine-tuned version of NAMAA-Space/AraModernBert-Base-V1.0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6959
• Accuracy: 0.8738
• Macro F1: 0.7566
• F1 Incoherent: 0.5877
• Precision Incoherent: 0.6381
• Recall Incoherent: 0.5447

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: 3e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• 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
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1	F1 Incoherent	Precision Incoherent	Recall Incoherent
6.5997	1.0	375	0.3938	0.8362	0.5991	0.2907	0.5102	0.2033
4.6902	2.0	750	0.4099	0.8550	0.5997	0.2800	0.7778	0.1707
3.4927	3.0	1125	0.4294	0.8725	0.7313	0.5366	0.6707	0.4472
0.9152	4.0	1500	0.6959	0.8738	0.7566	0.5877	0.6381	0.5447

Framework versions

• Transformers 5.6.2
• Pytorch 2.4.1+cu124
• Datasets 4.8.4
• Tokenizers 0.22.2