kavg/LiLT-RE-SIN

checkpoints

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:

• Precision: 0.2809
• Recall: 0.5051
• F1: 0.3610
• Loss: 1.6168

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: 1e-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
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 8000

Training results

Training Loss	Epoch	Step	F1	Validation Loss	Precision	Recall
0.1546	41.67	500	0	0.2482	0	0
0.1674	83.33	1000	0	0.2477	0	0
0.1368	125.0	1500	0.1502	0.2256	0.1975	0.1212
0.0727	166.67	2000	0.2732	0.3218	0.2091	0.3939
0.0718	208.33	2500	0.3385	0.3518	0.2579	0.4924
0.0612	250.0	3000	0.3371	0.5235	0.2555	0.4949
0.0504	291.67	3500	0.3353	0.5280	0.2536	0.4949
0.0418	333.33	4000	0.3476	0.6919	0.2657	0.5025
0.0308	375.0	4500	0.3490	0.7819	0.2613	0.5253
0.039	416.67	5000	0.3463	1.0291	0.2627	0.5076
0.0301	458.33	5500	0.3443	1.1661	0.2626	0.5
0.0245	500.0	6000	0.3414	1.2341	0.2642	0.4823
0.0347	541.67	6500	0.3389	1.4114	0.2605	0.4848
0.0327	583.33	7000	0.3422	1.4326	0.2683	0.4722
0.0117	625.0	7500	0.3670	1.6092	0.2899	0.5
0.0255	666.67	8000	0.3607	1.6141	0.2805	0.5051

Framework versions

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