LiLT-SER-DE-SIN

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

• Loss: 1.2864
• Precision: 0.7494
• Recall: 0.7734
• F1: 0.7612
• Accuracy: 0.8555

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
0.0053	21.74	500	0.9346	0.6644	0.7315	0.6964	0.8395
0.0026	43.48	1000	0.9918	0.7216	0.7278	0.7247	0.8587
0.0016	65.22	1500	0.9133	0.7098	0.7291	0.7193	0.8603
0.0001	86.96	2000	1.0041	0.7580	0.7328	0.7451	0.8536
0.0002	108.7	2500	1.1795	0.7312	0.7438	0.7375	0.8461
0.0001	130.43	3000	1.1417	0.7169	0.7328	0.7247	0.8642
0.0	152.17	3500	1.1740	0.7336	0.7562	0.7447	0.8453
0.0	173.91	4000	1.0535	0.7350	0.7414	0.7382	0.8635
0.0	195.65	4500	1.2769	0.7443	0.7241	0.7341	0.8502
0.0	217.39	5000	1.2235	0.7148	0.7438	0.7290	0.8363
0.0004	239.13	5500	1.2500	0.7376	0.7685	0.7527	0.8569
0.0	260.87	6000	1.2864	0.7494	0.7734	0.7612	0.8555
0.0	282.61	6500	1.1766	0.7649	0.7451	0.7548	0.8589
0.0	304.35	7000	1.3060	0.7231	0.7365	0.7297	0.8479
0.0	326.09	7500	1.1780	0.7093	0.7451	0.7267	0.8534
0.0	347.83	8000	1.2882	0.7337	0.75	0.7418	0.8614
0.0	369.57	8500	1.2833	0.7436	0.75	0.7468	0.8644
0.0	391.3	9000	1.4372	0.7372	0.7463	0.7417	0.8522
0.0	413.04	9500	1.4223	0.7382	0.75	0.7440	0.8513
0.0	434.78	10000	1.3686	0.7454	0.75	0.7477	0.8554

Framework versions

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