UNER_subword_tk_en_lora_alpha_128_drop_0.3_rank_64_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: 0.0651
• Precision: 0.7743
• Recall: 0.8168
• F1: 0.7950
• Accuracy: 0.9840

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	0.0794	0.6286	0.7692	0.6918	0.9751
0.1471	2.0	784	0.0628	0.7327	0.7547	0.7435	0.9814
0.0508	3.0	1176	0.0591	0.7203	0.7971	0.7568	0.9821
0.0409	4.0	1568	0.0571	0.7071	0.8147	0.7571	0.9819
0.0409	5.0	1960	0.0591	0.7139	0.8188	0.7628	0.9811
0.0345	6.0	2352	0.0556	0.7346	0.8023	0.7669	0.9827
0.031	7.0	2744	0.0598	0.7289	0.8209	0.7722	0.9819
0.0282	8.0	3136	0.0585	0.7671	0.8219	0.7936	0.9843
0.0247	9.0	3528	0.0567	0.7635	0.8219	0.7916	0.9840
0.0247	10.0	3920	0.0606	0.7830	0.7992	0.7910	0.9841
0.0225	11.0	4312	0.0567	0.7759	0.8137	0.7943	0.9849
0.0204	12.0	4704	0.0626	0.7724	0.8043	0.7880	0.9841
0.0195	13.0	5096	0.0600	0.7783	0.8106	0.7941	0.9849
0.0195	14.0	5488	0.0607	0.7671	0.8116	0.7887	0.9837
0.0184	15.0	5880	0.0629	0.7671	0.8116	0.7887	0.9837
0.0171	16.0	6272	0.0628	0.7767	0.8209	0.7982	0.9843
0.0155	17.0	6664	0.0631	0.7670	0.8075	0.7867	0.9841
0.0154	18.0	7056	0.0658	0.7673	0.8157	0.7908	0.9840
0.0154	19.0	7448	0.0651	0.7649	0.8219	0.7924	0.9841
0.0149	20.0	7840	0.0651	0.7743	0.8168	0.7950	0.9840

Framework versions

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