UNER_subword_tk_en_lora_alpha_64_drop_0.3_rank_32_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: 0.0607
• Precision: 0.7736
• Recall: 0.8240
• F1: 0.7980
• Accuracy: 0.9841

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	0.0899	0.5755	0.7143	0.6374	0.9740
0.1782	2.0	784	0.0651	0.6961	0.7919	0.7409	0.9799
0.0539	3.0	1176	0.0664	0.7144	0.8209	0.7640	0.9815
0.0435	4.0	1568	0.0581	0.7170	0.8209	0.7654	0.9821
0.0435	5.0	1960	0.0584	0.7321	0.8261	0.7763	0.9820
0.0385	6.0	2352	0.0571	0.7409	0.8230	0.7798	0.9827
0.0342	7.0	2744	0.0580	0.7433	0.8333	0.7857	0.9829
0.0313	8.0	3136	0.0578	0.7744	0.8282	0.8004	0.9846
0.0295	9.0	3528	0.0566	0.7588	0.8271	0.7915	0.9835
0.0295	10.0	3920	0.0564	0.7756	0.8302	0.8020	0.9848
0.0272	11.0	4312	0.0557	0.7597	0.8344	0.7953	0.9835
0.0256	12.0	4704	0.0585	0.7787	0.8157	0.7968	0.9841
0.0248	13.0	5096	0.0574	0.7812	0.8240	0.8020	0.9845
0.0248	14.0	5488	0.0577	0.7604	0.8344	0.7957	0.9836
0.023	15.0	5880	0.0583	0.7812	0.8282	0.8040	0.9845
0.0222	16.0	6272	0.0595	0.7733	0.8333	0.8022	0.9841
0.0205	17.0	6664	0.0603	0.7755	0.8261	0.8	0.9839
0.0207	18.0	7056	0.0605	0.7744	0.8282	0.8004	0.9840
0.0207	19.0	7448	0.0611	0.7770	0.8333	0.8042	0.9842
0.0203	20.0	7840	0.0607	0.7736	0.8240	0.7980	0.9841

Framework versions

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