File size: 3,476 Bytes
5a797b4
 
 
cb127fb
5a797b4
 
 
 
 
 
 
 
 
 
 
 
 
cb127fb
5a797b4
cb127fb
 
5a797b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
tags:
- image-classification
- generated_from_trainer
metrics:
- accuracy
model-index:
- name: exper3_mesum5
  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. -->

# exper3_mesum5

This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the sudo-s/herbier_mesuem5 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.6366
- Accuracy: 0.8367

## 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: 0.0002
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 8
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 3.895         | 0.23  | 100  | 3.8276          | 0.1935   |
| 3.1174        | 0.47  | 200  | 3.1217          | 0.3107   |
| 2.6           | 0.7   | 300  | 2.5399          | 0.4207   |
| 2.256         | 0.93  | 400  | 2.1767          | 0.5160   |
| 1.5441        | 1.16  | 500  | 1.8086          | 0.5852   |
| 1.3834        | 1.4   | 600  | 1.5565          | 0.6325   |
| 1.1995        | 1.63  | 700  | 1.3339          | 0.6763   |
| 1.0845        | 1.86  | 800  | 1.3299          | 0.6533   |
| 0.6472        | 2.09  | 900  | 1.0679          | 0.7219   |
| 0.5948        | 2.33  | 1000 | 1.0286          | 0.7124   |
| 0.5565        | 2.56  | 1100 | 0.9595          | 0.7284   |
| 0.4879        | 2.79  | 1200 | 0.8915          | 0.7420   |
| 0.2816        | 3.02  | 1300 | 0.8159          | 0.7763   |
| 0.2412        | 3.26  | 1400 | 0.7766          | 0.7911   |
| 0.2015        | 3.49  | 1500 | 0.7850          | 0.7828   |
| 0.274         | 3.72  | 1600 | 0.7361          | 0.7935   |
| 0.1244        | 3.95  | 1700 | 0.7299          | 0.7911   |
| 0.0794        | 4.19  | 1800 | 0.7441          | 0.7846   |
| 0.0915        | 4.42  | 1900 | 0.7614          | 0.7941   |
| 0.0817        | 4.65  | 2000 | 0.7310          | 0.8012   |
| 0.0561        | 4.88  | 2100 | 0.7222          | 0.8065   |
| 0.0165        | 5.12  | 2200 | 0.7515          | 0.8059   |
| 0.0168        | 5.35  | 2300 | 0.6687          | 0.8213   |
| 0.0212        | 5.58  | 2400 | 0.6671          | 0.8249   |
| 0.0389        | 5.81  | 2500 | 0.6893          | 0.8278   |
| 0.0087        | 6.05  | 2600 | 0.6839          | 0.8260   |
| 0.0087        | 6.28  | 2700 | 0.6412          | 0.8320   |
| 0.0077        | 6.51  | 2800 | 0.6366          | 0.8367   |
| 0.0065        | 6.74  | 2900 | 0.6697          | 0.8272   |
| 0.0061        | 6.98  | 3000 | 0.6510          | 0.8349   |
| 0.0185        | 7.21  | 3100 | 0.6452          | 0.8367   |
| 0.0059        | 7.44  | 3200 | 0.6426          | 0.8379   |
| 0.0062        | 7.67  | 3300 | 0.6398          | 0.8379   |
| 0.0315        | 7.91  | 3400 | 0.6397          | 0.8385   |


### Framework versions

- Transformers 4.20.1
- Pytorch 1.12.0+cu113
- Datasets 2.3.2
- Tokenizers 0.12.1