File size: 2,370 Bytes
806cc83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
---
license: cc-by-nc-sa-4.0
base_model: microsoft/layoutlmv3-base
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
- accuracy
model-index:
- name: primo_test
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# primo_test

This model is a fine-tuned version of [microsoft/layoutlmv3-base](https://huggingface.co/microsoft/layoutlmv3-base) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0105
- Precision: 0.9744
- Recall: 0.9902
- F1: 0.9822
- Accuracy: 0.9979

## 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: 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: 1000

### Training results

| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| No log        | 0.03  | 100  | 0.4646          | 0.8275    | 0.8656 | 0.8461 | 0.9306   |
| No log        | 0.06  | 200  | 0.0824          | 0.9614    | 0.9722 | 0.9667 | 0.9948   |
| No log        | 0.08  | 300  | 0.0363          | 0.9622    | 0.9859 | 0.9739 | 0.9951   |
| No log        | 0.11  | 400  | 0.0182          | 0.9756    | 0.9912 | 0.9833 | 0.9980   |
| 0.3067        | 0.14  | 500  | 0.0217          | 0.9578    | 0.9813 | 0.9694 | 0.9960   |
| 0.3067        | 0.17  | 600  | 0.0106          | 0.9913    | 0.9946 | 0.9929 | 0.9988   |
| 0.3067        | 0.19  | 700  | 0.0121          | 0.9733    | 0.9894 | 0.9812 | 0.9977   |
| 0.3067        | 0.22  | 800  | 0.0126          | 0.9699    | 0.9881 | 0.9789 | 0.9975   |
| 0.3067        | 0.25  | 900  | 0.0098          | 0.9778    | 0.9915 | 0.9846 | 0.9982   |
| 0.0105        | 0.28  | 1000 | 0.0105          | 0.9744    | 0.9902 | 0.9822 | 0.9979   |


### Framework versions

- Transformers 4.36.1
- Pytorch 2.1.1+cu121
- Datasets 2.15.0
- Tokenizers 0.15.0