lilt-form-read / README.md
romin23's picture
End of training
43661b7
metadata
license: mit
tags:
  - generated_from_trainer
datasets:
  - funsd-layoutlmv3
model-index:
  - name: lilt-form-read
    results: []

lilt-form-read

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: 1.7208
  • Answer: {'precision': 0.8635321100917431, 'recall': 0.9216646266829865, 'f1': 0.8916518650088809, 'number': 817}
  • Header: {'precision': 0.6813186813186813, 'recall': 0.5210084033613446, 'f1': 0.5904761904761905, 'number': 119}
  • Question: {'precision': 0.9005424954792043, 'recall': 0.924791086350975, 'f1': 0.9125057260650481, 'number': 1077}
  • Overall Precision: 0.8753
  • Overall Recall: 0.8997
  • Overall F1: 0.8873
  • Overall Accuracy: 0.8077

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
0.4555 10.53 200 0.9514 {'precision': 0.8207440811724915, 'recall': 0.8910648714810282, 'f1': 0.8544600938967137, 'number': 817} {'precision': 0.6233766233766234, 'recall': 0.40336134453781514, 'f1': 0.48979591836734687, 'number': 119} {'precision': 0.8611825192802056, 'recall': 0.9331476323119777, 'f1': 0.8957219251336899, 'number': 1077} 0.8358 0.8847 0.8596 0.7991
0.0457 21.05 400 1.4096 {'precision': 0.8654088050314466, 'recall': 0.8421052631578947, 'f1': 0.8535980148883374, 'number': 817} {'precision': 0.5833333333333334, 'recall': 0.5294117647058824, 'f1': 0.5550660792951542, 'number': 119} {'precision': 0.8606837606837607, 'recall': 0.9350046425255338, 'f1': 0.8963061860258122, 'number': 1077} 0.8480 0.8733 0.8605 0.7914
0.0144 31.58 600 1.4435 {'precision': 0.8720095693779905, 'recall': 0.8922888616891065, 'f1': 0.8820326678765881, 'number': 817} {'precision': 0.6428571428571429, 'recall': 0.5294117647058824, 'f1': 0.5806451612903226, 'number': 119} {'precision': 0.8682581786030061, 'recall': 0.9117920148560817, 'f1': 0.8894927536231884, 'number': 1077} 0.8591 0.8813 0.8700 0.8033
0.008 42.11 800 1.5197 {'precision': 0.8660287081339713, 'recall': 0.8861689106487148, 'f1': 0.8759830611010284, 'number': 817} {'precision': 0.5798319327731093, 'recall': 0.5798319327731093, 'f1': 0.5798319327731093, 'number': 119} {'precision': 0.8838248436103664, 'recall': 0.9182915506035283, 'f1': 0.9007285974499089, 'number': 1077} 0.8592 0.8852 0.8720 0.7921
0.0039 52.63 1000 1.4373 {'precision': 0.8733727810650888, 'recall': 0.9033047735618115, 'f1': 0.888086642599278, 'number': 817} {'precision': 0.6019417475728155, 'recall': 0.5210084033613446, 'f1': 0.5585585585585585, 'number': 119} {'precision': 0.8854351687388987, 'recall': 0.9257195914577531, 'f1': 0.9051293690422152, 'number': 1077} 0.8664 0.8927 0.8794 0.8096
0.0028 63.16 1200 1.7146 {'precision': 0.8490351872871736, 'recall': 0.9155446756425949, 'f1': 0.8810365135453475, 'number': 817} {'precision': 0.6941176470588235, 'recall': 0.4957983193277311, 'f1': 0.5784313725490197, 'number': 119} {'precision': 0.8852313167259787, 'recall': 0.9238625812441968, 'f1': 0.9041344843253067, 'number': 1077} 0.8622 0.8952 0.8784 0.7971
0.0022 73.68 1400 1.5638 {'precision': 0.8608893956670467, 'recall': 0.9241126070991432, 'f1': 0.8913813459268004, 'number': 817} {'precision': 0.6565656565656566, 'recall': 0.5462184873949579, 'f1': 0.5963302752293578, 'number': 119} {'precision': 0.8993536472760849, 'recall': 0.904363974001857, 'f1': 0.9018518518518519, 'number': 1077} 0.8713 0.8912 0.8811 0.8051
0.0009 84.21 1600 1.7113 {'precision': 0.8682080924855491, 'recall': 0.9192166462668299, 'f1': 0.8929845422116528, 'number': 817} {'precision': 0.65625, 'recall': 0.5294117647058824, 'f1': 0.586046511627907, 'number': 119} {'precision': 0.9085027726432532, 'recall': 0.9127205199628597, 'f1': 0.9106067623899953, 'number': 1077} 0.8796 0.8927 0.8861 0.8039
0.0009 94.74 1800 1.6397 {'precision': 0.8767942583732058, 'recall': 0.8971848225214198, 'f1': 0.8868723532970357, 'number': 817} {'precision': 0.6274509803921569, 'recall': 0.5378151260504201, 'f1': 0.579185520361991, 'number': 119} {'precision': 0.898458748866727, 'recall': 0.9201485608170845, 'f1': 0.9091743119266055, 'number': 1077} 0.8760 0.8882 0.8821 0.8042
0.0004 105.26 2000 1.7362 {'precision': 0.8690614136732329, 'recall': 0.9179926560587516, 'f1': 0.8928571428571428, 'number': 817} {'precision': 0.6458333333333334, 'recall': 0.5210084033613446, 'f1': 0.5767441860465117, 'number': 119} {'precision': 0.8928892889288929, 'recall': 0.9210770659238626, 'f1': 0.9067641681901281, 'number': 1077} 0.8715 0.8962 0.8837 0.8040
0.0003 115.79 2200 1.7208 {'precision': 0.8635321100917431, 'recall': 0.9216646266829865, 'f1': 0.8916518650088809, 'number': 817} {'precision': 0.6813186813186813, 'recall': 0.5210084033613446, 'f1': 0.5904761904761905, 'number': 119} {'precision': 0.9005424954792043, 'recall': 0.924791086350975, 'f1': 0.9125057260650481, 'number': 1077} 0.8753 0.8997 0.8873 0.8077
0.0002 126.32 2400 1.7281 {'precision': 0.8819362455726092, 'recall': 0.9143206854345165, 'f1': 0.8978365384615384, 'number': 817} {'precision': 0.6631578947368421, 'recall': 0.5294117647058824, 'f1': 0.5887850467289719, 'number': 119} {'precision': 0.8917710196779964, 'recall': 0.9257195914577531, 'f1': 0.9084282460136676, 'number': 1077} 0.8772 0.8977 0.8873 0.8060

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu116
  • Datasets 2.7.1
  • Tokenizers 0.13.2