kavg/LiLT-RE-EN

checkpoints

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

• Precision: 0.3264
• Recall: 0.4864
• F1: 0.3907
• Loss: 0.4377

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: 1e-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
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 5000

Training results

Training Loss	Epoch	Step	F1	Validation Loss	Precision	Recall
0.1604	26.32	500	0	0.1513	0	0
0.1012	52.63	1000	0.0098	0.0786	0.5	0.0049
0.0994	78.95	1500	0.2518	0.1847	0.3729	0.1901
0.0694	105.26	2000	0.3499	0.1926	0.3667	0.3346
0.0771	131.58	2500	0.3856	0.3295	0.3450	0.4370
0.0565	157.89	3000	0.3865	0.4137	0.3293	0.4679
0.0411	184.21	3500	0.3808	0.3624	0.3252	0.4593
0.0463	210.53	4000	0.3832	0.5089	0.3221	0.4728
0.0414	236.84	4500	0.3911	0.6137	0.3305	0.4790
0.036	263.16	5000	0.3910	0.4428	0.3275	0.4852

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1