LiLT-SER-ZH-SIN

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

• Loss: 1.2037
• Precision: 0.7417
• Recall: 0.7709
• F1: 0.7560
• Accuracy: 0.8558

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
0.0013	21.74	500	0.9018	0.6843	0.7475	0.7145	0.8599
0.012	43.48	1000	1.0791	0.7115	0.7623	0.7360	0.8561
0.0002	65.22	1500	1.0060	0.7360	0.7623	0.7489	0.8565
0.03	86.96	2000	1.1521	0.7282	0.6700	0.6979	0.8313
0.0013	108.7	2500	1.1517	0.7240	0.7463	0.7350	0.8579
0.0016	130.43	3000	0.9393	0.7319	0.7697	0.7503	0.8732
0.0021	152.17	3500	0.9972	0.7249	0.7562	0.7402	0.8635
0.0001	173.91	4000	1.0485	0.7049	0.7796	0.7404	0.8583
0.0002	195.65	4500	1.0827	0.7055	0.7315	0.7183	0.8433
0.0	217.39	5000	1.0528	0.7354	0.7599	0.7474	0.8586
0.0001	239.13	5500	1.1183	0.7001	0.7131	0.7065	0.8465
0.0002	260.87	6000	1.1749	0.7231	0.7685	0.7451	0.8520
0.0	282.61	6500	1.1206	0.7315	0.7685	0.7495	0.8611
0.0	304.35	7000	1.2037	0.7417	0.7709	0.7560	0.8558
0.0	326.09	7500	1.3737	0.7391	0.75	0.7445	0.8513
0.0	347.83	8000	1.2926	0.7221	0.7648	0.7428	0.8475
0.0	369.57	8500	1.4108	0.6966	0.7549	0.7246	0.8293
0.0	391.3	9000	1.4346	0.7222	0.7586	0.7399	0.8303
0.0	413.04	9500	1.4146	0.7225	0.7599	0.7407	0.8363
0.0	434.78	10000	1.4097	0.7121	0.7586	0.7346	0.8346

Framework versions

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