metadata

license: cc-by-nc-sa-4.0
base_model: microsoft/layoutlmv3-base
tags:
  - generated_from_trainer
  - Document Layout
  - LayoutLMv3
datasets:
  - funsd-layoutlmv3
metrics:
  - f1
  - accuracy
  - recall
  - precision
model-index:
  - name: layoutlmv3-base-fine_tuned-FUNSD_dataset
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: funsd-layoutlmv3
          type: funsd-layoutlmv3
          config: funsd
          split: test
          args: funsd
        metrics:
          - name: Precision
            type: precision
            value: 0.8978890525282278
          - name: Recall
            type: recall
            value: 0.9085941381023348
          - name: F1
            type: f1
            value: 0.9032098765432099
          - name: Accuracy
            type: accuracy
            value: 0.8461904195887318
language:
  - en

layoutlmv3-base-fine_tuned-FUNSD_dataset

This model is a fine-tuned version of microsoft/layoutlmv3-base on the funsd-layoutlmv3 dataset.

It achieves the following results on the evaluation set:

Loss: 0.2956
Precision: 0.8979
Recall: 0.9086
F1: 0.9032
Accuracy: 0.8462

Model description

For more information on how it was created, check out the following link: https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/blob/main/Document%20AI/Document%20Layout/LayoutLMv3%20with%20FUNSD/Fine%20tuning%20%26%20Evaluation%20-%20LayoutLMv3%20with%20FUNSD.ipynb

Intended uses & limitations

This model is intended to demonstrate my ability to solve a complex problem using technology.

Training and evaluation data

Dataset Source: https://huggingface.co/datasets/nielsr/funsd-layoutlmv3

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
training_steps: 2000

Training results

Train Loss	Epoch	Step	Valid. Loss	Precision	Recall	F1	Accuracy
0.2149	1.33	100	0.2402	0.7469	0.8212	0.7823	0.7758
0.1466	2.67	200	0.1869	0.8161	0.8838	0.8486	0.8273
0.1122	4.0	300	0.1902	0.8538	0.8997	0.8761	0.8316
0.0757	5.33	400	0.1857	0.8354	0.8927	0.8631	0.8349
0.0427	6.67	500	0.2091	0.8792	0.8897	0.8844	0.8446
0.0495	8.0	600	0.2235	0.8825	0.9031	0.8927	0.8370
0.0369	9.33	700	0.2532	0.8826	0.9146	0.8983	0.8349
0.0329	10.67	800	0.2576	0.8829	0.8992	0.8910	0.8474
0.0229	12.0	900	0.2579	0.8827	0.8937	0.8882	0.8443
0.0219	13.33	1000	0.2710	0.8710	0.8987	0.8846	0.8347
0.0191	14.67	1100	0.2582	0.8889	0.9061	0.8974	0.8454
0.0179	16.0	1200	0.2646	0.8870	0.9006	0.8938	0.8356
0.0135	17.33	1300	0.2798	0.8949	0.9180	0.9063	0.8512
0.007	18.67	1400	0.2944	0.8988	0.9091	0.9039	0.8455
0.0064	20.0	1500	0.2822	0.8938	0.9071	0.9004	0.8452
0.0089	21.33	1600	0.3003	0.8941	0.9101	0.9020	0.8484
0.0099	22.67	1700	0.3008	0.8942	0.9071	0.9006	0.8439
0.0069	24.0	1800	0.2965	0.8942	0.9071	0.9006	0.8386
0.0048	25.33	1900	0.2973	0.9027	0.9076	0.9051	0.8501
0.0069	26.67	2000	0.2956	0.8979	0.9086	0.9032	0.8462

Framework versions

Transformers 4.33.2
Pytorch 2.0.1
Datasets 2.14.5
Tokenizers 0.13.3