File size: 2,591 Bytes
9751501
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- marsyas/gtzan
metrics:
- accuracy
model-index:
- name: distilhubert-finetuned-gtzan
  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. -->

# distilhubert-finetuned-gtzan

This model is a fine-tuned version of [ntu-spml/distilhubert](https://huggingface.co/ntu-spml/distilhubert) on the GTZAN dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0379
- Accuracy: 0.81

## 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
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 20

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 2.0307        | 1.0   | 113  | 2.0561          | 0.41     |
| 1.4208        | 2.0   | 226  | 1.4850          | 0.63     |
| 1.1959        | 3.0   | 339  | 1.0617          | 0.66     |
| 0.6929        | 4.0   | 452  | 0.8228          | 0.74     |
| 0.5104        | 5.0   | 565  | 0.6969          | 0.77     |
| 0.4735        | 6.0   | 678  | 0.7412          | 0.79     |
| 0.2185        | 7.0   | 791  | 0.6586          | 0.76     |
| 0.3087        | 8.0   | 904  | 0.8234          | 0.78     |
| 0.1066        | 9.0   | 1017 | 0.8210          | 0.8      |
| 0.0841        | 10.0  | 1130 | 1.0040          | 0.8      |
| 0.0387        | 11.0  | 1243 | 0.9195          | 0.81     |
| 0.0091        | 12.0  | 1356 | 0.9208          | 0.82     |
| 0.006         | 13.0  | 1469 | 0.9190          | 0.81     |
| 0.0051        | 14.0  | 1582 | 0.9796          | 0.8      |
| 0.0038        | 15.0  | 1695 | 0.9823          | 0.8      |
| 0.0035        | 16.0  | 1808 | 1.0252          | 0.8      |
| 0.0032        | 17.0  | 1921 | 1.0172          | 0.8      |
| 0.0032        | 18.0  | 2034 | 1.0433          | 0.81     |
| 0.0029        | 19.0  | 2147 | 1.0577          | 0.81     |
| 0.0029        | 20.0  | 2260 | 1.0379          | 0.81     |


### Framework versions

- Transformers 4.30.2
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3