UNER_subword_tk_en_lora_alpha_512_drop_0.3_rank_256_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: 0.0920
• Precision: 0.802
• Recall: 0.8302
• F1: 0.8159
• Accuracy: 0.9850

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	0.0788	0.6623	0.7816	0.7170	0.9762
0.1163	2.0	784	0.0638	0.7512	0.7940	0.7720	0.9822
0.0451	3.0	1176	0.0626	0.7405	0.8240	0.7800	0.9827
0.0331	4.0	1568	0.0597	0.7341	0.8147	0.7723	0.9824
0.0331	5.0	1960	0.0617	0.7648	0.8313	0.7966	0.9845
0.0267	6.0	2352	0.0635	0.7725	0.8261	0.7984	0.9848
0.0215	7.0	2744	0.0693	0.7786	0.8302	0.8036	0.9843
0.0177	8.0	3136	0.0768	0.7866	0.8126	0.7994	0.9845
0.0149	9.0	3528	0.0675	0.7968	0.8364	0.8162	0.9851
0.0149	10.0	3920	0.0708	0.8100	0.8251	0.8174	0.9855
0.0131	11.0	4312	0.0723	0.7913	0.8282	0.8093	0.9850
0.0107	12.0	4704	0.0799	0.7900	0.8219	0.8057	0.9846
0.0098	13.0	5096	0.0834	0.8069	0.8178	0.8123	0.9845
0.0098	14.0	5488	0.0817	0.7960	0.8199	0.8078	0.9844
0.0082	15.0	5880	0.0864	0.8022	0.8230	0.8125	0.9848
0.0073	16.0	6272	0.0906	0.7984	0.8199	0.8090	0.9848
0.0059	17.0	6664	0.0899	0.8077	0.8261	0.8168	0.9850
0.006	18.0	7056	0.0914	0.7940	0.8261	0.8097	0.9851
0.006	19.0	7448	0.0914	0.7939	0.8333	0.8131	0.9847
0.0052	20.0	7840	0.0920	0.802	0.8302	0.8159	0.9850

Framework versions

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