LiLT-SER-IT

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

• Loss: 2.5355
• Precision: 0.7262
• Recall: 0.7927
• F1: 0.7580
• Accuracy: 0.7687

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.0696	7.46	500	1.0876	0.6322	0.6517	0.6418	0.7584
0.0576	14.93	1000	1.3989	0.6712	0.7601	0.7129	0.7601
0.0096	22.39	1500	1.8059	0.6774	0.7639	0.7181	0.7662
0.0092	29.85	2000	2.0416	0.7266	0.7334	0.7300	0.7652
0.0003	37.31	2500	2.0467	0.7166	0.7539	0.7348	0.7628
0.0013	44.78	3000	2.0159	0.7027	0.7821	0.7403	0.7638
0.0013	52.24	3500	2.2751	0.6961	0.7728	0.7325	0.7575
0.0002	59.7	4000	2.2084	0.7236	0.7563	0.7396	0.7723
0.0002	67.16	4500	2.1843	0.7048	0.7701	0.7360	0.7581
0.0001	74.63	5000	2.2483	0.7366	0.7745	0.7551	0.7770
0.0001	82.09	5500	2.2685	0.7171	0.7752	0.7451	0.7677
0.0005	89.55	6000	2.2877	0.7180	0.7821	0.7487	0.7692
0.0001	97.01	6500	2.2574	0.7308	0.7725	0.7511	0.7721
0.0	104.48	7000	2.4696	0.7255	0.7862	0.7546	0.7660
0.0	111.94	7500	2.3996	0.7140	0.7917	0.7509	0.7725
0.0	119.4	8000	2.4592	0.7261	0.7852	0.7545	0.7665
0.0	126.87	8500	2.4129	0.7336	0.7900	0.7607	0.7718
0.0	134.33	9000	2.5367	0.7316	0.7896	0.7595	0.7666
0.0	141.79	9500	2.5327	0.7278	0.7900	0.7576	0.7663
0.0	149.25	10000	2.5355	0.7262	0.7927	0.7580	0.7687

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.1