LiLT-SER-ES

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: 2.5588
• Precision: 0.6719
• Recall: 0.6734
• F1: 0.6726
• Accuracy: 0.7463

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
0.2279	8.2	500	0.6790	0.5205	1.2508	0.4589	0.6012
0.032	16.39	1000	0.6936	0.5885	1.9637	0.6321	0.5505
0.0073	24.59	1500	0.7351	0.6175	1.6711	0.5795	0.6608
0.0479	32.79	2000	0.7405	0.6422	1.8259	0.6265	0.6586
0.0666	40.98	2500	0.7424	0.6349	1.8343	0.5937	0.6824
0.0006	49.18	3000	0.7475	0.6536	2.0575	0.6512	0.6559
0.0084	57.38	3500	0.7138	0.6415	2.4488	0.6758	0.6106
0.0002	65.57	4000	0.7571	0.6468	1.9641	0.6406	0.6532
0.0005	73.77	4500	2.2976	0.6699	0.6429	0.6561	0.7413
0.0003	81.97	5000	2.1562	0.6287	0.6653	0.6465	0.7468
0.0007	90.16	5500	2.2806	0.6435	0.6689	0.6560	0.7435
0.0002	98.36	6000	2.0508	0.6294	0.6734	0.6506	0.7538
0.0	106.56	6500	2.2626	0.6602	0.6765	0.6683	0.7498
0.0	114.75	7000	2.3467	0.6687	0.6492	0.6588	0.7409
0.0	122.95	7500	2.4430	0.6773	0.6734	0.6754	0.7447
0.0	131.15	8000	2.3653	0.6643	0.6765	0.6704	0.7476
0.0	139.34	8500	2.2903	0.6567	0.6824	0.6693	0.7498
0.0	147.54	9000	2.4458	0.6536	0.6824	0.6677	0.7440
0.0	155.74	9500	2.5953	0.6703	0.6685	0.6694	0.7423
0.0	163.93	10000	2.5588	0.6719	0.6734	0.6726	0.7463

Framework versions

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