File size: 5,439 Bytes
5e09fbc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
base_model: facebook/bart-large-xsum
tags:
- generated_from_trainer
metrics:
- rouge
model-index:
- name: text_shortening_model_v37
  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. -->

# text_shortening_model_v37

This model is a fine-tuned version of [facebook/bart-large-xsum](https://huggingface.co/facebook/bart-large-xsum) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 2.9472
- Rouge1: 0.4923
- Rouge2: 0.2809
- Rougel: 0.4462
- Rougelsum: 0.4468
- Bert precision: 0.8731
- Bert recall: 0.8773
- Average word count: 9.1021
- Max word count: 15
- Min word count: 5
- Average token count: 16.8198
- % shortened texts with length > 12: 8.7087

## 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: 0.0003
- 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
- num_epochs: 15

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Bert precision | Bert recall | Average word count | Max word count | Min word count | Average token count | % shortened texts with length > 12 |
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:---------:|:--------------:|:-----------:|:------------------:|:--------------:|:--------------:|:-------------------:|:----------------------------------:|
| 1.5911        | 1.0   | 73   | 1.8586          | 0.4823 | 0.2756 | 0.4416 | 0.4423    | 0.8661         | 0.8758      | 8.9399             | 21             | 4              | 16.9489             | 7.8078                             |
| 0.9246        | 2.0   | 146  | 2.2274          | 0.4039 | 0.2049 | 0.3771 | 0.3764    | 0.8526         | 0.855       | 8.0991             | 13             | 4              | 14.6006             | 0.6006                             |
| 0.7574        | 3.0   | 219  | 1.8752          | 0.4463 | 0.2263 | 0.4072 | 0.4071    | 0.8629         | 0.8654      | 8.3934             | 14             | 5              | 14.3303             | 3.003                              |
| 0.6131        | 4.0   | 292  | 1.8338          | 0.4896 | 0.2691 | 0.4451 | 0.4456    | 0.8747         | 0.8711      | 7.982              | 13             | 4              | 13.9249             | 0.3003                             |
| 0.4422        | 5.0   | 365  | 1.8257          | 0.492  | 0.2727 | 0.4499 | 0.4504    | 0.8734         | 0.875       | 8.5165             | 16             | 5              | 14.4595             | 3.003                              |
| 0.4227        | 6.0   | 438  | 2.1249          | 0.4666 | 0.2475 | 0.418  | 0.4178    | 0.8657         | 0.8697      | 9.3874             | 16             | 4              | 16.9399             | 8.4084                             |
| 0.3714        | 7.0   | 511  | 2.1010          | 0.4838 | 0.274  | 0.436  | 0.4364    | 0.869          | 0.8754      | 9.4264             | 16             | 5              | 14.9369             | 9.009                              |
| 0.2638        | 8.0   | 584  | 2.0803          | 0.489  | 0.2799 | 0.4404 | 0.4404    | 0.8701         | 0.8751      | 8.976              | 15             | 4              | 15.5736             | 8.4084                             |
| 0.2103        | 9.0   | 657  | 2.1093          | 0.4888 | 0.2722 | 0.4381 | 0.438     | 0.872          | 0.8751      | 9.1952             | 16             | 5              | 16.7447             | 9.9099                             |
| 0.1475        | 10.0  | 730  | 2.3159          | 0.4684 | 0.2597 | 0.4243 | 0.4244    | 0.8632         | 0.8721      | 9.4234             | 15             | 5              | 16.8288             | 11.7117                            |
| 0.122         | 11.0  | 803  | 2.4090          | 0.4845 | 0.2729 | 0.4421 | 0.4427    | 0.8721         | 0.8748      | 8.8018             | 16             | 5              | 16.4264             | 5.7057                             |
| 0.0915        | 12.0  | 876  | 2.6598          | 0.4838 | 0.2691 | 0.4376 | 0.437     | 0.8698         | 0.8742      | 9.1652             | 16             | 5              | 16.9009             | 10.2102                            |
| 0.073         | 13.0  | 949  | 2.5266          | 0.4973 | 0.2861 | 0.4479 | 0.4495    | 0.8743         | 0.8776      | 9.0631             | 16             | 5              | 16.5796             | 8.4084                             |
| 0.0526        | 14.0  | 1022 | 2.7673          | 0.4955 | 0.2821 | 0.4464 | 0.4463    | 0.8716         | 0.8791      | 9.4685             | 16             | 5              | 17.2012             | 10.5105                            |
| 0.042         | 15.0  | 1095 | 2.9472          | 0.4923 | 0.2809 | 0.4462 | 0.4468    | 0.8731         | 0.8773      | 9.1021             | 15             | 5              | 16.8198             | 8.7087                             |


### Framework versions

- Transformers 4.33.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3