langid-ner-multilingual-bert-gn-base-cased

This model is a fine-tuned version of mmaguero/multilingual-bert-gn-base-cased on the task 1 and task 2 of GUA-SPA@IberLEF 2023 shared task dataset. It achieves the following results on the evaluation set:

• Loss: 0.5980
• Precision: 0.7389
• Recall: 0.7464
• F1: 0.7426
• Accuracy: 0.8673

Model description

More information needed

Intended uses & limitations

• NER (PER, LOC, ORG)
• Token-based language identification (es, gn, mix, foreign)

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: 16
• eval_batch_size: 16
• 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	72	1.1288	0.5004	0.4849	0.4925	0.7053
No log	2.0	144	0.7476	0.5986	0.5881	0.5933	0.7979
No log	3.0	216	0.6203	0.6329	0.6334	0.6331	0.8244
No log	4.0	288	0.5784	0.6413	0.6644	0.6527	0.8387
No log	5.0	360	0.5487	0.6626	0.6871	0.6746	0.8464
No log	6.0	432	0.5396	0.6929	0.7097	0.7012	0.8581
0.692	7.0	504	0.5446	0.6918	0.7156	0.7035	0.8555
0.692	8.0	576	0.5528	0.6932	0.7240	0.7082	0.8602
0.692	9.0	648	0.5614	0.6987	0.7257	0.7119	0.8602
0.692	10.0	720	0.5751	0.7071	0.7290	0.7179	0.8598
0.692	11.0	792	0.5865	0.6991	0.7232	0.7109	0.8602
0.692	12.0	864	0.5910	0.7102	0.7341	0.7219	0.8648
0.692	13.0	936	0.6068	0.7131	0.7383	0.7255	0.8602
0.161	14.0	1008	0.6168	0.7066	0.7374	0.7217	0.8635
0.161	15.0	1080	0.6177	0.7061	0.7357	0.7206	0.8628
0.161	16.0	1152	0.6280	0.7130	0.7399	0.7262	0.8632
0.161	17.0	1224	0.6293	0.7071	0.7391	0.7227	0.8628
0.161	18.0	1296	0.6330	0.7104	0.7408	0.7253	0.8638
0.161	19.0	1368	0.6365	0.7088	0.7391	0.7236	0.8625
0.161	20.0	1440	0.6367	0.7093	0.7391	0.7239	0.8615

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1