lilt-en-funsd

This model is a fine-tuned version of SCUT-DLVCLab/lilt-roberta-en-base on the funsd-layoutlmv3 dataset. It achieves the following results on the evaluation set:

  • Loss: 1.5815
  • Answer: {'precision': 0.8604118993135011, 'recall': 0.9204406364749081, 'f1': 0.8894145476049675, 'number': 817}
  • Header: {'precision': 0.6330275229357798, 'recall': 0.5798319327731093, 'f1': 0.6052631578947367, 'number': 119}
  • Question: {'precision': 0.9101851851851852, 'recall': 0.9127205199628597, 'f1': 0.9114510894761243, 'number': 1077}
  • Overall Precision: 0.8745
  • Overall Recall: 0.8962
  • Overall F1: 0.8852
  • Overall Accuracy: 0.8209

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: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 2500
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Answer Header Question Overall Precision Overall Recall Overall F1 Overall Accuracy
0.4131 10.53 200 0.9920 {'precision': 0.7944444444444444, 'recall': 0.8751529987760098, 'f1': 0.8328479906814211, 'number': 817} {'precision': 0.5267857142857143, 'recall': 0.4957983193277311, 'f1': 0.5108225108225107, 'number': 119} {'precision': 0.8690265486725663, 'recall': 0.9117920148560817, 'f1': 0.8898957861350248, 'number': 1077} 0.8198 0.8723 0.8452 0.7912
0.0453 21.05 400 1.3055 {'precision': 0.8215077605321508, 'recall': 0.9069767441860465, 'f1': 0.8621291448516578, 'number': 817} {'precision': 0.5961538461538461, 'recall': 0.5210084033613446, 'f1': 0.5560538116591929, 'number': 119} {'precision': 0.8818755635707844, 'recall': 0.9080779944289693, 'f1': 0.8947849954254347, 'number': 1077} 0.8421 0.8847 0.8629 0.7971
0.0129 31.58 600 1.6559 {'precision': 0.8261826182618262, 'recall': 0.9192166462668299, 'f1': 0.8702201622247971, 'number': 817} {'precision': 0.4957983193277311, 'recall': 0.4957983193277311, 'f1': 0.4957983193277311, 'number': 119} {'precision': 0.9050814956855225, 'recall': 0.8765088207985144, 'f1': 0.8905660377358492, 'number': 1077} 0.8469 0.8713 0.8590 0.7952
0.0083 42.11 800 1.6136 {'precision': 0.8760529482551144, 'recall': 0.8910648714810282, 'f1': 0.883495145631068, 'number': 817} {'precision': 0.6145833333333334, 'recall': 0.4957983193277311, 'f1': 0.5488372093023256, 'number': 119} {'precision': 0.8963922294172063, 'recall': 0.8997214484679665, 'f1': 0.8980537534754401, 'number': 1077} 0.8745 0.8723 0.8734 0.8060
0.0058 52.63 1000 1.6826 {'precision': 0.8553386911595867, 'recall': 0.9118727050183598, 'f1': 0.8827014218009479, 'number': 817} {'precision': 0.6355140186915887, 'recall': 0.5714285714285714, 'f1': 0.6017699115044248, 'number': 119} {'precision': 0.8902991840435177, 'recall': 0.9117920148560817, 'f1': 0.9009174311926607, 'number': 1077} 0.8626 0.8917 0.8769 0.7928
0.0027 63.16 1200 1.5511 {'precision': 0.8640661938534279, 'recall': 0.8947368421052632, 'f1': 0.8791340950090198, 'number': 817} {'precision': 0.576, 'recall': 0.6050420168067226, 'f1': 0.5901639344262294, 'number': 119} {'precision': 0.8985374771480804, 'recall': 0.9127205199628597, 'f1': 0.9055734684477199, 'number': 1077} 0.8649 0.8872 0.8759 0.8110
0.0014 73.68 1400 1.5130 {'precision': 0.8801452784503632, 'recall': 0.8898408812729498, 'f1': 0.8849665246500303, 'number': 817} {'precision': 0.6213592233009708, 'recall': 0.5378151260504201, 'f1': 0.5765765765765765, 'number': 119} {'precision': 0.8748906386701663, 'recall': 0.9285051067780873, 'f1': 0.900900900900901, 'number': 1077} 0.8644 0.8897 0.8769 0.8092
0.001 84.21 1600 1.5433 {'precision': 0.8373893805309734, 'recall': 0.9265605875152999, 'f1': 0.8797210923881464, 'number': 817} {'precision': 0.6033057851239669, 'recall': 0.6134453781512605, 'f1': 0.6083333333333334, 'number': 119} {'precision': 0.9138257575757576, 'recall': 0.8960074280408542, 'f1': 0.9048288795124239, 'number': 1077} 0.8626 0.8917 0.8769 0.8139
0.0006 94.74 1800 1.5585 {'precision': 0.8500576701268743, 'recall': 0.9020807833537332, 'f1': 0.8752969121140143, 'number': 817} {'precision': 0.6371681415929203, 'recall': 0.6050420168067226, 'f1': 0.6206896551724138, 'number': 119} {'precision': 0.8933454876937101, 'recall': 0.9099350046425255, 'f1': 0.9015639374425023, 'number': 1077} 0.8613 0.8887 0.8748 0.8197
0.0003 105.26 2000 1.5719 {'precision': 0.8505096262740657, 'recall': 0.9192166462668299, 'f1': 0.8835294117647059, 'number': 817} {'precision': 0.6605504587155964, 'recall': 0.6050420168067226, 'f1': 0.6315789473684209, 'number': 119} {'precision': 0.9113805970149254, 'recall': 0.9071494893221913, 'f1': 0.9092601209865054, 'number': 1077} 0.8721 0.8942 0.8830 0.8246
0.0004 115.79 2200 1.5578 {'precision': 0.8554913294797688, 'recall': 0.9057527539779682, 'f1': 0.8799048751486326, 'number': 817} {'precision': 0.6283185840707964, 'recall': 0.5966386554621849, 'f1': 0.6120689655172413, 'number': 119} {'precision': 0.9059907834101383, 'recall': 0.9127205199628597, 'f1': 0.9093432007400555, 'number': 1077} 0.8696 0.8912 0.8803 0.8194
0.0003 126.32 2400 1.5815 {'precision': 0.8604118993135011, 'recall': 0.9204406364749081, 'f1': 0.8894145476049675, 'number': 817} {'precision': 0.6330275229357798, 'recall': 0.5798319327731093, 'f1': 0.6052631578947367, 'number': 119} {'precision': 0.9101851851851852, 'recall': 0.9127205199628597, 'f1': 0.9114510894761243, 'number': 1077} 0.8745 0.8962 0.8852 0.8209

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.15.0
  • Tokenizers 0.15.0
Downloads last month
19
Safetensors
Model size
130M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for Ifyouknowthenyouknow/lilt-en-funsd

Finetuned
(46)
this model