File size: 2,688 Bytes
40503f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
base_model: dima806/music_genres_classification
tags:
- generated_from_trainer
datasets:
- marsyas/gtzan
metrics:
- accuracy
model-index:
- name: music_genres_classification-finetuned-gtzan
  results:
  - task:
      name: Audio Classification
      type: audio-classification
    dataset:
      name: GTZAN
      type: marsyas/gtzan
      config: all
      split: train
      args: all
    metrics:
    - name: Accuracy
      type: accuracy
      value: 0.87
---

<!-- 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. -->

# music_genres_classification-finetuned-gtzan

This model is a fine-tuned version of [dima806/music_genres_classification](https://huggingface.co/dima806/music_genres_classification) on the GTZAN dataset.
It achieves the following results on the evaluation set:
- Loss: 1.0028
- Accuracy: 0.87

## 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: 5
- eval_batch_size: 5
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.12
- num_epochs: 15
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 1.5758        | 1.0   | 180  | 1.5756          | 0.52     |
| 1.079         | 2.0   | 360  | 1.2787          | 0.6      |
| 1.186         | 3.0   | 540  | 1.3863          | 0.58     |
| 0.9181        | 4.0   | 720  | 1.3967          | 0.64     |
| 0.4356        | 5.0   | 900  | 1.2449          | 0.67     |
| 0.4013        | 6.0   | 1080 | 1.2714          | 0.71     |
| 0.5518        | 7.0   | 1260 | 0.8282          | 0.8      |
| 0.4808        | 8.0   | 1440 | 1.3598          | 0.75     |
| 0.3608        | 9.0   | 1620 | 1.1908          | 0.8      |
| 0.181         | 10.0  | 1800 | 0.9824          | 0.83     |
| 0.0553        | 11.0  | 1980 | 1.0336          | 0.84     |
| 0.2445        | 12.0  | 2160 | 1.1085          | 0.83     |
| 0.0103        | 13.0  | 2340 | 1.1288          | 0.84     |
| 0.2437        | 14.0  | 2520 | 1.0183          | 0.85     |
| 0.0921        | 15.0  | 2700 | 1.0028          | 0.87     |


### Framework versions

- Transformers 4.39.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2