File size: 2,492 Bytes
2a96cf6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: mit
base_model: ayameRushia/bert-base-indonesian-1.5G-sentiment-analysis-smsa
tags:
- generated_from_trainer
metrics:
- accuracy
- precision
- recall
- f1
model-index:
- name: best_berita_bert_model_fold_3
  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. -->

# best_berita_bert_model_fold_3

This model is a fine-tuned version of [ayameRushia/bert-base-indonesian-1.5G-sentiment-analysis-smsa](https://huggingface.co/ayameRushia/bert-base-indonesian-1.5G-sentiment-analysis-smsa) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1732
- Accuracy: 0.9808
- Precision: 0.9809
- Recall: 0.9811
- F1: 0.9809

## 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: 5e-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: 10

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1     |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:|
| 0.5469        | 1.0   | 601  | 0.2814          | 0.9409   | 0.9416    | 0.9414 | 0.9410 |
| 0.21          | 2.0   | 1202 | 0.1697          | 0.9600   | 0.9602    | 0.9605 | 0.9600 |
| 0.1002        | 3.0   | 1803 | 0.2227          | 0.9667   | 0.9674    | 0.9673 | 0.9666 |
| 0.0847        | 4.0   | 2404 | 0.2771          | 0.9584   | 0.9599    | 0.9592 | 0.9581 |
| 0.029         | 5.0   | 3005 | 0.1732          | 0.9808   | 0.9809    | 0.9811 | 0.9809 |
| 0.0095        | 6.0   | 3606 | 0.2415          | 0.9734   | 0.9737    | 0.9738 | 0.9733 |
| 0.0134        | 7.0   | 4207 | 0.2048          | 0.9767   | 0.9769    | 0.9771 | 0.9766 |
| 0.0001        | 8.0   | 4808 | 0.2916          | 0.9692   | 0.9697    | 0.9698 | 0.9691 |
| 0.0039        | 9.0   | 5409 | 0.2201          | 0.9784   | 0.9786    | 0.9787 | 0.9784 |
| 0.0           | 10.0  | 6010 | 0.2293          | 0.9742   | 0.9745    | 0.9746 | 0.9742 |


### Framework versions

- Transformers 4.41.2
- Pytorch 2.1.2
- Datasets 2.19.2
- Tokenizers 0.19.1