multilingual-google/mt5-base-hausa-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: 0.1053
• Precision: 0.9485
• Recall: 0.8978
• F1: 0.9224
• Accuracy: 0.9827

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	301	0.2665	0.7902	0.7083	0.7470	0.9413
0.9875	2.0	602	0.2028	0.8688	0.8055	0.8360	0.9588
0.9875	3.0	903	0.1779	0.8880	0.8120	0.8483	0.9613
0.2051	4.0	1204	0.2090	0.8725	0.8506	0.8614	0.9638
0.157	5.0	1505	0.1443	0.9138	0.8397	0.8752	0.9677
0.157	6.0	1806	0.1434	0.9186	0.8571	0.8867	0.9706
0.1285	7.0	2107	0.1401	0.9102	0.8680	0.8886	0.9713
0.1285	8.0	2408	0.1483	0.9079	0.8693	0.8882	0.9706
0.1101	9.0	2709	0.1349	0.9274	0.8635	0.8943	0.9732
0.0985	10.0	3010	0.1217	0.9232	0.8667	0.8941	0.9734
0.0985	11.0	3311	0.1501	0.9190	0.8764	0.8972	0.9734
0.0879	12.0	3612	0.1474	0.9047	0.8802	0.8923	0.9727
0.0879	13.0	3913	0.1323	0.9245	0.8828	0.9032	0.9746
0.0786	14.0	4214	0.1496	0.9155	0.8796	0.8972	0.9743
0.0721	15.0	4515	0.1489	0.9238	0.8815	0.9021	0.9754
0.0721	16.0	4816	0.1399	0.9188	0.8892	0.9038	0.9759
0.0652	17.0	5117	0.1377	0.9149	0.8925	0.9035	0.9763
0.0652	18.0	5418	0.1476	0.9158	0.8970	0.9063	0.9772
0.0539	19.0	5719	0.1403	0.9189	0.8976	0.9081	0.9774
0.0534	20.0	6020	0.1404	0.9190	0.8989	0.9089	0.9776

Framework versions

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