multilingual-google/mt5-base-luganda-ner-v1

This model is a fine-tuned version of google/mt5-base on the Beijuka/Multilingual_PII_NER_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.3334
• Precision: 0.8281
• Recall: 0.5568
• F1: 0.6659
• Accuracy: 0.9215

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
No log	1.0	261	1.1064	0.1333	0.0021	0.0041	0.8000
1.773	2.0	522	1.1335	0.0	0.0	0.0	0.7945
1.773	3.0	783	1.0490	0.1818	0.0021	0.0041	0.7955
1.0922	4.0	1044	0.9485	0.3810	0.0251	0.0471	0.8030
1.0922	5.0	1305	0.8640	0.4933	0.0387	0.0717	0.8093
0.9396	6.0	1566	0.6608	0.5948	0.1902	0.2882	0.8404
0.9396	7.0	1827	0.5730	0.6781	0.2685	0.3847	0.8569
0.6952	8.0	2088	0.4691	0.6998	0.3605	0.4759	0.8768
0.6952	9.0	2349	0.4007	0.7271	0.4399	0.5482	0.9004
0.5088	10.0	2610	0.4192	0.6621	0.5037	0.5721	0.8947
0.5088	11.0	2871	0.4036	0.6886	0.5361	0.6028	0.9005
0.4013	12.0	3132	0.3698	0.7103	0.5381	0.6124	0.9093
0.4013	13.0	3393	0.3491	0.7279	0.5423	0.6216	0.9137
0.351	14.0	3654	0.3207	0.8056	0.5413	0.6475	0.9242
0.351	15.0	3915	0.3423	0.7697	0.5413	0.6356	0.9197
0.308	16.0	4176	0.3359	0.7783	0.5465	0.6421	0.9220
0.308	17.0	4437	0.3334	0.7713	0.5496	0.6419	0.9216

Framework versions

• Transformers 4.55.4
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4