kavg/LiLT-RE-JA

checkpoints

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:

• Precision: 0.4372
• Recall: 0.6574
• F1: 0.5252
• Loss: 0.0001

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: 10000

Training results

Training Loss	Epoch	Step	F1	Validation Loss	Precision	Recall
0.1954	20.0	500	0	0.4094	0	0
0.1588	40.0	1000	0.1420	0.3055	0.3587	0.0886
0.1182	60.0	1500	0.4253	0.1384	0.3810	0.4812
0.0477	80.0	2000	0.4764	0.0216	0.3949	0.6002
0.069	100.0	2500	0.5198	0.0115	0.4564	0.6038
0.0355	120.0	3000	0.5161	0.0018	0.4271	0.6521
0.0268	140.0	3500	0.5254	0.0016	0.4395	0.6530
0.0123	160.0	4000	0.5264	0.0015	0.4382	0.6592
0.0039	180.0	4500	0.5353	0.0011	0.4510	0.6583
0.0139	200.0	5000	0.5390	0.0011	0.4533	0.6646
0.001	220.0	5500	0.5430	0.0042	0.4620	0.6583
0.01	240.0	6000	0.5347	0.0013	0.4531	0.6521
0.0065	260.0	6500	0.5404	0.0001	0.4540	0.6673
0.0046	280.0	7000	0.5252	0.0001	0.4372	0.6574
0.002	300.0	7500	0.5365	0.0007	0.4474	0.6699
0.0002	320.0	8000	0.5393	0.0002	0.4546	0.6628
0.0008	340.0	8500	0.5412	0.0002	0.4569	0.6637
0.0024	360.0	9000	0.4677	0.6601	0.5475	0.0002
0.0001	380.0	9500	0.4560	0.6673	0.5418	0.0002
0.002	400.0	10000	0.4594	0.6628	0.5427	0.0003

Framework versions

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