multilingual-afroxlmr-large-ner-masakhaner-v2

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: 0.2439
• Precision: 0.9331
• Recall: 0.9297
• F1: 0.9314
• Accuracy: 0.9740

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
0.2022	1.0	1260	0.3005	0.8695	0.8567	0.8630	0.9519
0.1243	2.0	2520	0.2468	0.8609	0.8762	0.8685	0.9594
0.0887	3.0	3780	0.2360	0.8709	0.9029	0.8866	0.9631
0.0618	4.0	5040	0.2571	0.8843	0.9020	0.8930	0.9633
0.2729	5.0	6300	0.3104	0.9051	0.7911	0.8443	0.9420
0.0529	6.0	7560	0.2458	0.8958	0.8735	0.8845	0.9611
0.0372	7.0	8820	0.2298	0.9107	0.9129	0.9118	0.9686
0.0252	8.0	10080	0.2377	0.8972	0.9113	0.9042	0.9684
0.0172	9.0	11340	0.2759	0.9145	0.8889	0.9015	0.9662
0.0187	10.0	12600	0.2502	0.9102	0.9223	0.9162	0.9726
0.0136	11.0	13860	0.2539	0.9120	0.9214	0.9167	0.9720
0.0091	12.0	15120	0.2402	0.9017	0.9348	0.9179	0.9708
0.0062	13.0	16380	0.2836	0.9009	0.9298	0.9151	0.9706
0.0056	14.0	17640	0.2745	0.9050	0.9253	0.9151	0.9707
0.0023	15.0	18900	0.3044	0.9143	0.9231	0.9187	0.9722
0.0039	16.0	20160	0.2707	0.9127	0.9332	0.9228	0.9727
0.0025	17.0	21420	0.2990	0.9183	0.9239	0.9211	0.9723
0.0013	18.0	22680	0.2724	0.9195	0.9314	0.9254	0.9733
0.0011	19.0	23940	0.3120	0.9133	0.9317	0.9224	0.9720
0.0008	20.0	25200	0.3058	0.9133	0.9308	0.9220	0.9721

Framework versions

• Transformers 4.55.4
• Pytorch 2.8.0.dev20250319+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4