LiLT-SER-JA-SIN

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

• Loss: 1.1113
• Precision: 0.7378
• Recall: 0.7660
• F1: 0.7517
• Accuracy: 0.8794

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
0.0009	21.74	500	0.8785	0.6584	0.7217	0.6886	0.8505
0.0031	43.48	1000	1.0637	0.7309	0.7291	0.7300	0.8533
0.0046	65.22	1500	0.9166	0.7219	0.7512	0.7363	0.8729
0.0002	86.96	2000	1.0366	0.7212	0.7389	0.7299	0.8721
0.0	108.7	2500	1.0535	0.7191	0.7377	0.7283	0.8662
0.0006	130.43	3000	1.1869	0.7409	0.7291	0.7349	0.8495
0.005	152.17	3500	1.2062	0.7356	0.7401	0.7379	0.8627
0.0002	173.91	4000	1.2067	0.7011	0.7192	0.7100	0.8451
0.0002	195.65	4500	1.1819	0.7290	0.7389	0.7339	0.8578
0.0	217.39	5000	1.1699	0.7463	0.75	0.7482	0.8632
0.0	239.13	5500	1.1548	0.7267	0.7599	0.7429	0.8637
0.0	260.87	6000	1.1867	0.7227	0.7574	0.7396	0.8651
0.0	282.61	6500	1.1614	0.7222	0.7525	0.7370	0.8721
0.0	304.35	7000	1.1884	0.7146	0.7648	0.7388	0.8681
0.0	326.09	7500	1.2186	0.6975	0.7438	0.7199	0.8582
0.0001	347.83	8000	1.0423	0.7313	0.7709	0.7506	0.8754
0.0	369.57	8500	1.1254	0.7278	0.7574	0.7423	0.8705
0.0	391.3	9000	1.1113	0.7378	0.7660	0.7517	0.8794
0.0	413.04	9500	1.1517	0.7424	0.7562	0.7492	0.8732
0.0	434.78	10000	1.1568	0.7413	0.7586	0.7498	0.8726

Framework versions

• Transformers 4.39.1
• Pytorch 2.1.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.1