luganda-ner-v1

This model is a fine-tuned version of xlm-roberta-base on the lg-ner dataset. It achieves the following results on the evaluation set:

• Loss: 0.3188
• Precision: 0.7988
• Recall: 0.8026
• F1: 0.8007
• Accuracy: 0.9452

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	261	0.2915	0.7456	0.6891	0.7162	0.9240
0.2284	2.0	522	0.2965	0.7393	0.7314	0.7353	0.9294
0.2284	3.0	783	0.2830	0.7426	0.7576	0.7500	0.9271
0.1426	4.0	1044	0.2710	0.7935	0.7690	0.7810	0.9387
0.1426	5.0	1305	0.2805	0.8087	0.7636	0.7855	0.9389
0.0881	6.0	1566	0.2992	0.7734	0.7884	0.7808	0.9404
0.0881	7.0	1827	0.2746	0.8109	0.7864	0.7985	0.9457
0.0582	8.0	2088	0.3149	0.7753	0.7925	0.7838	0.9400
0.0582	9.0	2349	0.3179	0.7940	0.7945	0.7942	0.9440
0.0403	10.0	2610	0.3188	0.7988	0.8026	0.8007	0.9452

Framework versions

• Transformers 4.27.4
• Pytorch 1.13.1+cu116
• Datasets 2.11.0
• Tokenizers 0.13.2