LiLT-SER-IT-SIN

This model is a fine-tuned version of kavg/LiLT-SER-IT on the xfun dataset. It achieves the following results on the evaluation set:

• Loss: 1.2031
• Precision: 0.7651
• Recall: 0.7783
• F1: 0.7717
• Accuracy: 0.8705

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
0.0301	21.74	500	1.0148	0.7146	0.7586	0.7360	0.8470
0.0058	43.48	1000	0.9498	0.7121	0.7401	0.7258	0.8566
0.0008	65.22	1500	1.0385	0.7310	0.7833	0.7562	0.8559
0.0004	86.96	2000	1.2165	0.7484	0.7032	0.7251	0.8512
0.0017	108.7	2500	1.0999	0.7252	0.7734	0.7485	0.8726
0.0018	130.43	3000	1.1872	0.7293	0.7697	0.7490	0.8564
0.0001	152.17	3500	1.2632	0.7386	0.7377	0.7381	0.8457
0.0008	173.91	4000	1.0687	0.7337	0.7635	0.7483	0.8691
0.0	195.65	4500	1.0346	0.7205	0.7746	0.7466	0.8684
0.0	217.39	5000	1.1440	0.7158	0.7537	0.7343	0.8686
0.0	239.13	5500	1.3391	0.7690	0.7586	0.7638	0.8578
0.0	260.87	6000	1.0498	0.7482	0.7722	0.7600	0.8761
0.0	282.61	6500	1.0602	0.7301	0.7894	0.7586	0.8787
0.0	304.35	7000	1.1634	0.7355	0.7328	0.7341	0.8613
0.0	326.09	7500	1.1705	0.7680	0.7746	0.7713	0.8754
0.0	347.83	8000	1.2455	0.7616	0.7709	0.7662	0.8687
0.0	369.57	8500	1.2259	0.7327	0.7562	0.7442	0.8665
0.0	391.3	9000	1.1737	0.7577	0.7857	0.7715	0.8690
0.0	413.04	9500	1.2174	0.7636	0.7796	0.7715	0.8704
0.0	434.78	10000	1.2031	0.7651	0.7783	0.7717	0.8705

Framework versions

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