LiLT-SER-SIN

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:

• Loss: 1.1967
• Precision: 0.7058
• Recall: 0.7475
• F1: 0.7261
• Accuracy: 0.8621

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	Accuracy	F1	Validation Loss	Precision	Recall
0.0739	21.74	500	0.8268	0.5620	0.7143	0.5	0.6416
0.0509	43.48	1000	0.8324	0.5839	0.8499	0.5348	0.6429
0.0004	65.22	1500	0.8398	0.6521	0.9889	0.6256	0.6810
0.0004	86.96	2000	0.8461	0.6678	1.0577	0.6251	0.7167
0.003	108.7	2500	0.8561	0.6929	1.0734	0.6532	0.7377
0.0006	130.43	3000	0.8569	0.6924	1.1114	0.6686	0.7180
0.0022	152.17	3500	0.8245	0.6749	1.4184	0.6774	0.6724
0.0001	173.91	4000	0.8502	0.6937	1.0524	0.6546	0.7377
0.001	195.65	4500	0.8493	0.6900	1.1949	0.6663	0.7155
0.0001	217.39	5000	0.8460	0.6885	1.1462	0.6790	0.6983
0.0001	239.13	5500	0.8641	0.6970	1.1296	0.6697	0.7266
0.0	260.87	6000	0.8529	0.7046	1.2585	0.6929	0.7167
0.0037	282.61	6500	0.8634	0.7139	1.2292	0.6917	0.7377
0.0	304.35	7000	0.8621	0.7261	1.1967	0.7058	0.7475
0.0	326.09	7500	0.8585	0.7230	1.2144	0.7089	0.7377
0.0	347.83	8000	0.8609	0.7180	1.2117	0.6918	0.7463
0.0	369.57	8500	0.8628	0.7135	1.1961	0.6755	0.7562
0.0	391.3	9000	0.8624	0.7220	1.2292	0.7059	0.7389
0.0	413.04	9500	0.8611	0.7262	1.2278	0.7071	0.7463
0.0	434.78	10000	0.8609	0.7242	1.2317	0.7056	0.7438

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2