LiLT-SER-FR-SIN

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

• Loss: 1.2426
• Precision: 0.7618
• Recall: 0.7956
• F1: 0.7783
• Accuracy: 0.8648

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
0.0057	21.74	500	0.8019	0.6884	0.7020	0.6951	0.8582
0.008	43.48	1000	1.0139	0.6963	0.7623	0.7278	0.8648
0.0006	65.22	1500	0.9878	0.7090	0.7562	0.7318	0.8592
0.0038	86.96	2000	1.2269	0.7104	0.7401	0.7250	0.8373
0.001	108.7	2500	0.9751	0.7276	0.7697	0.7481	0.8707
0.0004	130.43	3000	1.0918	0.7479	0.7672	0.7574	0.8538
0.0003	152.17	3500	1.0782	0.7102	0.7635	0.7359	0.8604
0.0	173.91	4000	1.0515	0.7402	0.7894	0.7640	0.8704
0.0001	195.65	4500	1.2154	0.7373	0.7709	0.7538	0.8419
0.0	217.39	5000	1.1026	0.7411	0.7722	0.7563	0.8642
0.0001	239.13	5500	1.0594	0.7262	0.7512	0.7385	0.8576
0.0	260.87	6000	1.1103	0.7377	0.7759	0.7563	0.8609
0.0	282.61	6500	1.1591	0.7267	0.7599	0.7429	0.8610
0.0	304.35	7000	1.2382	0.7574	0.7537	0.7556	0.8562
0.0	326.09	7500	1.2027	0.7485	0.7882	0.7678	0.8578
0.0001	347.83	8000	1.1492	0.7433	0.7808	0.7616	0.8659
0.0002	369.57	8500	1.1924	0.7570	0.7980	0.7770	0.8655
0.0	391.3	9000	1.2426	0.7618	0.7956	0.7783	0.8648
0.0	413.04	9500	1.3078	0.7620	0.7808	0.7713	0.8597
0.0	434.78	10000	1.3219	0.7639	0.7771	0.7705	0.8579

Framework versions

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