UNER_subword_tk_en_lora_alpha_256_drop_0.3_rank_128_seed_42

This model is a fine-tuned version of xlm-roberta-base on the universalner/universal_ner en_ewt dataset. It achieves the following results on the evaluation set:

• Loss: 0.0751
• Precision: 0.7770
• Recall: 0.8261
• F1: 0.8008
• Accuracy: 0.9842

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20.0

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	392	0.0746	0.6978	0.7578	0.7266	0.9775
0.1317	2.0	784	0.0618	0.7088	0.7888	0.7467	0.9807
0.0475	3.0	1176	0.0578	0.7483	0.8157	0.7806	0.9840
0.037	4.0	1568	0.0550	0.7439	0.8271	0.7833	0.9837
0.037	5.0	1960	0.0573	0.7468	0.8364	0.7891	0.9827
0.0305	6.0	2352	0.0581	0.7458	0.8230	0.7825	0.9833
0.0259	7.0	2744	0.0603	0.7683	0.8375	0.8014	0.9840
0.0237	8.0	3136	0.0622	0.7754	0.8219	0.7980	0.9843
0.0197	9.0	3528	0.0618	0.7759	0.8209	0.7978	0.9840
0.0197	10.0	3920	0.0664	0.7814	0.8178	0.7992	0.9845
0.0174	11.0	4312	0.0638	0.7751	0.8137	0.7939	0.9841
0.0152	12.0	4704	0.0678	0.7783	0.8251	0.8010	0.9845
0.0146	13.0	5096	0.0663	0.7871	0.8116	0.7992	0.9845
0.0146	14.0	5488	0.0678	0.7819	0.8313	0.8058	0.9849
0.0123	15.0	5880	0.0702	0.7862	0.8261	0.8057	0.9844
0.0115	16.0	6272	0.0727	0.7872	0.8271	0.8067	0.9846
0.0098	17.0	6664	0.0730	0.7952	0.8240	0.8094	0.9849
0.01	18.0	7056	0.0754	0.7891	0.8251	0.8067	0.9848
0.01	19.0	7448	0.0749	0.7706	0.8240	0.7964	0.9839
0.0091	20.0	7840	0.0751	0.7770	0.8261	0.8008	0.9842

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1