File size: 4,344 Bytes
1ad38bb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
---
license: cc-by-sa-4.0
tags:
- generated_from_trainer
metrics:
- accuracy
- precision
- recall
- f1
model-index:
- name: legal-bert-small
  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. -->

# legal-bert-small

This model is a fine-tuned version of [nlpaueb/legal-bert-small-uncased](https://huggingface.co/nlpaueb/legal-bert-small-uncased) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1837
- Accuracy: 0.9548
- Precision: 0.7491
- Recall: 0.7882
- F1: 0.7682
- Classification Report:               precision    recall  f1-score   support

         LOC       0.84      0.86      0.85      1668
        MISC       0.59      0.63      0.61       702
         ORG       0.64      0.67      0.66      1661
         PER       0.83      0.90      0.87      1617

   micro avg       0.75      0.79      0.77      5648
   macro avg       0.73      0.77      0.75      5648
weighted avg       0.75      0.79      0.77      5648


## 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: 2e-05
- train_batch_size: 16
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1     | Classification Report                                                                                                                                                                                                                                                                                                                                                                                                                              |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|
| 0.1443        | 1.0   | 434  | 0.1949          | 0.9462   | 0.6982    | 0.7466 | 0.7216 |               precision    recall  f1-score   support

         LOC       0.83      0.80      0.81      1668
        MISC       0.59      0.62      0.60       702
         ORG       0.56      0.58      0.57      1661
         PER       0.75      0.92      0.83      1617

   micro avg       0.70      0.75      0.72      5648
   macro avg       0.68      0.73      0.70      5648
weighted avg       0.70      0.75      0.72      5648
 |
| 0.071         | 2.0   | 868  | 0.1764          | 0.9551   | 0.7548    | 0.7702 | 0.7624 |               precision    recall  f1-score   support

         LOC       0.81      0.88      0.84      1668
        MISC       0.59      0.64      0.61       702
         ORG       0.68      0.59      0.63      1661
         PER       0.83      0.90      0.86      1617

   micro avg       0.75      0.77      0.76      5648
   macro avg       0.73      0.75      0.74      5648
weighted avg       0.75      0.77      0.76      5648
 |
| 0.0713        | 3.0   | 1302 | 0.1837          | 0.9548   | 0.7491    | 0.7882 | 0.7682 |               precision    recall  f1-score   support

         LOC       0.84      0.86      0.85      1668
        MISC       0.59      0.63      0.61       702
         ORG       0.64      0.67      0.66      1661
         PER       0.83      0.90      0.87      1617

   micro avg       0.75      0.79      0.77      5648
   macro avg       0.73      0.77      0.75      5648
weighted avg       0.75      0.79      0.77      5648
 |


### Framework versions

- Transformers 4.30.2
- Pytorch 2.0.0
- Datasets 2.1.0
- Tokenizers 0.13.3