File size: 2,411 Bytes
ba0f52a
 
 
 
 
 
 
9c38b18
 
 
ba0f52a
 
 
 
 
 
 
 
 
 
 
 
1ce9889
 
 
 
 
ba0f52a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1ce9889
9c38b18
 
ba0f52a
 
 
 
 
 
 
9c38b18
 
1ce9889
 
 
 
 
 
 
 
 
 
ba0f52a
 
 
 
 
 
 
 
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: Microsoft/Multilingual-MiniLM-L12-H384
tags:
- generated_from_trainer
metrics:
- accuracy
- f1
- precision
- recall
model-index:
- name: my-model-MiniLM-Area
  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. -->

# my-model-MiniLM-Area

This model is a fine-tuned version of [Microsoft/Multilingual-MiniLM-L12-H384](https://huggingface.co/Microsoft/Multilingual-MiniLM-L12-H384) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 1.4662
- Accuracy: 0.4676
- F1: 0.3710
- Precision: 0.3276
- Recall: 0.4676

## 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: 3e-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 | F1     | Precision | Recall |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:|
| 1.8319        | 1.0   | 81   | 1.7972          | 0.2446   | 0.0961 | 0.0598    | 0.2446 |
| 1.7578        | 2.0   | 162  | 1.7665          | 0.2446   | 0.0961 | 0.0598    | 0.2446 |
| 1.7311        | 3.0   | 243  | 1.7140          | 0.3885   | 0.2579 | 0.1991    | 0.3885 |
| 1.6502        | 4.0   | 324  | 1.6061          | 0.4173   | 0.2789 | 0.2142    | 0.4173 |
| 1.551         | 5.0   | 405  | 1.5444          | 0.4029   | 0.2666 | 0.1999    | 0.4029 |
| 1.4605        | 6.0   | 486  | 1.5607          | 0.4532   | 0.3470 | 0.2842    | 0.4532 |
| 1.3818        | 7.0   | 567  | 1.5001          | 0.4604   | 0.3552 | 0.2913    | 0.4604 |
| 1.3256        | 8.0   | 648  | 1.5020          | 0.4820   | 0.3847 | 0.3341    | 0.4820 |
| 1.3104        | 9.0   | 729  | 1.4776          | 0.4604   | 0.3644 | 0.3218    | 0.4604 |
| 1.2741        | 10.0  | 810  | 1.4662          | 0.4676   | 0.3710 | 0.3276    | 0.4676 |


### Framework versions

- Transformers 4.41.0
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1