File size: 3,371 Bytes
a665b82
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
base_model: facebook/bart-large-cnn
tags:
- generated_from_trainer
metrics:
- rouge
model-index:
- name: bart-large-cnn-finetuned-scope-summarization-train-test-split
  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. -->

# bart-large-cnn-finetuned-scope-summarization-train-test-split

This model is a fine-tuned version of [facebook/bart-large-cnn](https://huggingface.co./facebook/bart-large-cnn) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 1.2315
- Rouge1: 52.3537
- Rouge2: 31.6854
- Rougel: 36.6454
- Rougelsum: 50.8292

## 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: 5.6e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 20

### Training results

| Training Loss | Epoch | Step | Validation Loss | Rouge1  | Rouge2  | Rougel  | Rougelsum |
|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:-------:|:---------:|
| No log        | 1.0   | 25   | 0.6966          | 51.835  | 31.057  | 37.6234 | 50.2076   |
| 0.6673        | 2.0   | 50   | 0.6823          | 48.381  | 28.6493 | 37.1777 | 46.9784   |
| 0.5505        | 3.0   | 75   | 0.6825          | 51.1061 | 31.5147 | 38.5282 | 49.8741   |
| 0.5505        | 4.0   | 100  | 0.7131          | 51.0351 | 32.3268 | 39.7744 | 49.4893   |
| 0.4736        | 5.0   | 125  | 0.6975          | 52.9068 | 32.4415 | 39.5503 | 51.2993   |
| 0.4033        | 6.0   | 150  | 0.7925          | 51.3766 | 30.4233 | 37.7124 | 49.5155   |
| 0.3306        | 7.0   | 175  | 0.8079          | 52.2073 | 31.8487 | 38.6156 | 50.8166   |
| 0.3306        | 8.0   | 200  | 0.9168          | 51.6434 | 31.3338 | 37.4811 | 50.1527   |
| 0.256         | 9.0   | 225  | 0.9810          | 49.7984 | 30.3608 | 36.7693 | 48.7107   |
| 0.1823        | 10.0  | 250  | 0.9289          | 51.679  | 31.2458 | 36.4793 | 50.2032   |
| 0.1355        | 11.0  | 275  | 1.0269          | 52.0775 | 31.1824 | 37.5405 | 50.5995   |
| 0.1355        | 12.0  | 300  | 1.0736          | 51.3365 | 31.2121 | 38.37   | 50.0703   |
| 0.0974        | 13.0  | 325  | 1.0935          | 52.4146 | 32.5704 | 38.0578 | 51.424    |
| 0.0681        | 14.0  | 350  | 1.1100          | 51.5136 | 31.6307 | 38.5212 | 50.2267   |
| 0.0476        | 15.0  | 375  | 1.1507          | 51.9246 | 31.5588 | 36.8706 | 50.7219   |
| 0.0476        | 16.0  | 400  | 1.1667          | 53.7686 | 33.3238 | 38.145  | 52.2277   |
| 0.0336        | 17.0  | 425  | 1.1606          | 51.9682 | 31.4379 | 37.6764 | 50.8294   |
| 0.0232        | 18.0  | 450  | 1.1961          | 51.6253 | 31.6575 | 37.5128 | 50.406    |
| 0.0232        | 19.0  | 475  | 1.2162          | 51.7758 | 31.8239 | 36.3796 | 50.3009   |
| 0.0182        | 20.0  | 500  | 1.2315          | 52.3537 | 31.6854 | 36.6454 | 50.8292   |


### Framework versions

- Transformers 4.38.1
- Pytorch 2.2.1+cu121
- Datasets 2.17.1
- Tokenizers 0.15.2