LiLT-SER-PT-SIN

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

• Loss: 1.2074
• Precision: 0.7639
• Recall: 0.7771
• F1: 0.7705
• Accuracy: 0.8627

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
0.0124	21.74	500	0.8590	0.7403	0.8082	0.7381	0.7426
0.0017	43.48	1000	0.8301	0.7272	1.2608	0.75	0.7057
0.0004	65.22	1500	0.8694	0.7323	0.8843	0.7098	0.7562
0.0	86.96	2000	0.8617	0.7532	1.0638	0.7419	0.7648
0.0001	108.7	2500	0.8580	0.7674	1.1504	0.7689	0.7660
0.0006	130.43	3000	0.8677	0.7479	0.9865	0.7230	0.7746
0.0	152.17	3500	0.8617	0.7558	1.1492	0.7494	0.7623
0.0001	173.91	4000	0.8385	0.7590	1.3124	0.7485	0.7697
0.0055	195.65	4500	1.1331	0.7295	0.7869	0.7571	0.8479
0.0	217.39	5000	1.2061	0.7392	0.7611	0.7500	0.8500
0.0001	239.13	5500	1.2572	0.7253	0.7672	0.7457	0.8482
0.0	260.87	6000	1.3558	0.7494	0.7734	0.7612	0.8569
0.0	282.61	6500	1.4382	0.7598	0.7672	0.7635	0.8589
0.0	304.35	7000	1.4720	0.7537	0.7574	0.7555	0.8533
0.0	326.09	7500	1.3835	0.7524	0.7783	0.7651	0.8579
0.0	347.83	8000	1.2693	0.7534	0.7599	0.7566	0.8599
0.0	369.57	8500	1.2005	0.7417	0.7709	0.7560	0.8600
0.0	391.3	9000	1.2175	0.7560	0.7820	0.7688	0.8601
0.0	413.04	9500	1.2339	0.7556	0.7845	0.7698	0.8601
0.0	434.78	10000	1.2074	0.7639	0.7771	0.7705	0.8627

Framework versions

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