File size: 2,351 Bytes
1004f4a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
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: 0.7463
- Accuracy: 0.83

## 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
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 15
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 1.9408        | 1.0   | 113  | 1.9838          | 0.43     |
| 1.2842        | 2.0   | 226  | 1.2837          | 0.67     |
| 1.0008        | 3.0   | 339  | 0.9786          | 0.74     |
| 0.656         | 4.0   | 452  | 0.7425          | 0.83     |
| 0.39          | 5.0   | 565  | 0.5993          | 0.82     |
| 0.2612        | 6.0   | 678  | 0.6584          | 0.8      |
| 0.1779        | 7.0   | 791  | 0.5676          | 0.81     |
| 0.1512        | 8.0   | 904  | 0.9030          | 0.76     |
| 0.093         | 9.0   | 1017 | 0.7049          | 0.85     |
| 0.0355        | 10.0  | 1130 | 0.7865          | 0.82     |
| 0.0111        | 11.0  | 1243 | 0.7816          | 0.83     |
| 0.0088        | 12.0  | 1356 | 0.7861          | 0.82     |
| 0.0073        | 13.0  | 1469 | 0.7535          | 0.84     |
| 0.007         | 14.0  | 1582 | 0.7547          | 0.83     |
| 0.0063        | 15.0  | 1695 | 0.7463          | 0.83     |


### Framework versions

- Transformers 4.29.2
- Pytorch 1.13.1+cu117
- Datasets 2.12.0
- Tokenizers 0.13.3