File size: 2,884 Bytes
4a1e1ad
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d8947ce
 
 
 
 
 
 
4a1e1ad
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fdca2da
4a1e1ad
 
 
 
 
d8947ce
 
 
 
 
 
 
 
 
 
 
 
 
4a1e1ad
 
 
 
fdca2da
4a1e1ad
 
 
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
---
tags:
- generated_from_trainer
model-index:
- name: icdar23-entrydetector_plaintext_breaks
  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. -->

# icdar23-entrydetector_plaintext_breaks

This model is a fine-tuned version of [HueyNemud/das22-10-camembert_pretrained](https://huggingface.co/HueyNemud/das22-10-camembert_pretrained) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0120
- Ebegin: {'precision': 0.9901997738409348, 'recall': 0.9879654005265137, 'f1': 0.9890813253012049, 'number': 2659}
- Eend: {'precision': 0.9916824196597354, 'recall': 0.9801943198804185, 'f1': 0.9859049050930276, 'number': 2676}
- Overall Precision: 0.9909
- Overall Recall: 0.9841
- Overall F1: 0.9875
- Overall Accuracy: 0.9977

## 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: 0.0001
- 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: 7500

### Training results

| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1     | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
| No log        | 0.07  | 300  | 0.0318          | 0.9847    | 0.9761 | 0.9803 | 0.9966   |
| 0.1683        | 0.14  | 600  | 0.0164          | 0.9878    | 0.9890 | 0.9884 | 0.9978   |
| 0.1683        | 0.21  | 900  | 0.0146          | 0.9900    | 0.9853 | 0.9876 | 0.9976   |
| 0.0203        | 0.29  | 1200 | 0.0112          | 0.9862    | 0.9902 | 0.9882 | 0.9978   |
| 0.0123        | 0.36  | 1500 | 0.0089          | 0.9943    | 0.9878 | 0.9910 | 0.9983   |
| 0.0123        | 0.43  | 1800 | 0.0139          | 0.9970    | 0.9814 | 0.9891 | 0.9979   |
| 0.0109        | 0.5   | 2100 | 0.0101          | 0.9937    | 0.9882 | 0.9909 | 0.9982   |
| 0.0109        | 0.57  | 2400 | 0.0087          | 0.9949    | 0.9896 | 0.9922 | 0.9985   |
| 0.0092        | 0.64  | 2700 | 0.0081          | 0.9849    | 0.9919 | 0.9884 | 0.9978   |
| 0.0084        | 0.72  | 3000 | 0.0087          | 0.9937    | 0.9867 | 0.9902 | 0.9981   |
| 0.0084        | 0.79  | 3300 | 0.0089          | 0.9915    | 0.9889 | 0.9902 | 0.9981   |
| 0.0069        | 0.86  | 3600 | 0.0092          | 0.9899    | 0.9901 | 0.9900 | 0.9981   |
| 0.0069        | 0.93  | 3900 | 0.0097          | 0.9845    | 0.9915 | 0.9880 | 0.9977   |


### Framework versions

- Transformers 4.26.1
- Pytorch 1.13.1+cu116
- Datasets 2.9.0
- Tokenizers 0.13.2