pasha

This model is a fine-tuned version of microsoft/layoutlmv3-base on the nielsr/funsd-layoutlmv3 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0585
  • Precision: 0.9867
  • Recall: 0.9892
  • F1: 0.9879
  • Accuracy: 0.9906

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: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 1000

Training results

Training Loss Epoch Step Validation Loss Precision Recall F1 Accuracy
No log 2.13 100 0.2664 0.9534 0.9438 0.9486 0.9571
No log 4.26 200 0.1044 0.9756 0.9802 0.9779 0.9838
No log 6.38 300 0.0672 0.9853 0.9899 0.9876 0.9904
No log 8.51 400 0.0634 0.9824 0.9860 0.9842 0.9884
0.2958 10.64 500 0.0585 0.9867 0.9892 0.9879 0.9906
0.2958 12.77 600 0.0511 0.9889 0.9928 0.9908 0.9928
0.2958 14.89 700 0.0503 0.9871 0.9921 0.9896 0.9925
0.2958 17.02 800 0.0529 0.9860 0.9903 0.9881 0.9913
0.2958 19.15 900 0.0581 0.9842 0.9892 0.9867 0.9904
0.0256 21.28 1000 0.0571 0.9849 0.9888 0.9869 0.9901

Framework versions

  • Transformers 4.26.0.dev0
  • Pytorch 1.12.1
  • Datasets 2.6.1
  • Tokenizers 0.13.2
Downloads last month
13
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.

Dataset used to train mijungkim/pasha

Evaluation results