multilingual-afroxlmr-large-ner-masakhaner-ner-v1

This model is a fine-tuned version of masakhane/afroxlmr-large-ner-masakhaner-1.0_2.0 on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.2533
• Precision: 0.9394
• Recall: 0.9198
• F1: 0.9295
• Accuracy: 0.9736

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.201	1.0	1260	0.2656	0.8469	0.8570	0.8519	0.9540
0.1255	2.0	2520	0.2759	0.8698	0.8834	0.8765	0.9599
0.0913	3.0	3780	0.2277	0.8734	0.9144	0.8934	0.9628
0.0639	4.0	5040	0.2044	0.8839	0.9208	0.9020	0.9670
0.0535	5.0	6300	0.2354	0.8964	0.9022	0.8993	0.9664
0.0407	6.0	7560	0.2140	0.8913	0.9265	0.9086	0.9678
0.0265	7.0	8820	0.2100	0.9143	0.9074	0.9108	0.9687
0.0237	8.0	10080	0.2753	0.9038	0.9206	0.9121	0.9697
0.0175	9.0	11340	0.2501	0.9083	0.8979	0.9031	0.9688
0.0151	10.0	12600	0.2796	0.9009	0.9135	0.9071	0.9682
0.0123	11.0	13860	0.2927	0.9088	0.9197	0.9143	0.9698
0.0087	12.0	15120	0.2623	0.9023	0.9265	0.9143	0.9702
0.0058	13.0	16380	0.3155	0.9085	0.9165	0.9125	0.9684
0.0054	14.0	17640	0.2703	0.9118	0.9319	0.9218	0.9732
0.0048	15.0	18900	0.2875	0.9037	0.9312	0.9172	0.9715
0.0037	16.0	20160	0.2817	0.9079	0.9274	0.9176	0.9719
0.0012	17.0	21420	0.3166	0.8996	0.9339	0.9164	0.9711

Framework versions

• Transformers 4.55.2
• Pytorch 2.8.0.dev20250319+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4