File size: 2,742 Bytes
f53dc23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
96
---
license: apache-2.0
base_model: facebook/hubert-base-ls960
tags:
- generated_from_trainer
datasets:
- Emo-Codec/CREMA-D_synth
metrics:
- accuracy
- precision
- recall
- f1
model-index:
- name: hubert-base-ls960-tone-classification
  results:
  - task:
      name: Audio Classification
      type: audio-classification
    dataset:
      name: CREMA-D
      type: Emo-Codec/CREMA-D_synth
    metrics:
    - name: Accuracy
      type: accuracy
      value: 0.8016085790884718
    - name: Precision
      type: precision
      value: 0.8014677098753149
    - name: Recall
      type: recall
      value: 0.8016085790884718
    - name: F1
      type: f1
      value: 0.7989608760238184
---

<!-- 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. -->

# hubert-base-ls960-tone-classification

This model is a fine-tuned version of [facebook/hubert-base-ls960](https://huggingface.co/facebook/hubert-base-ls960) on the CREMA-D dataset.
It achieves the following results on the evaluation set:
- Loss: 0.7499
- Accuracy: 0.8016
- Precision: 0.8015
- Recall: 0.8016
- F1: 0.7990

## 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: 16
- eval_batch_size: 16
- 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: 8

### Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1     |
|:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:|
| 1.4326        | 1.0   | 442  | 1.2934          | 0.5147   | 0.5889    | 0.5147 | 0.4878 |
| 1.0447        | 2.0   | 884  | 0.8590          | 0.7051   | 0.7570    | 0.7051 | 0.7125 |
| 0.775         | 3.0   | 1326 | 0.7668          | 0.7426   | 0.7589    | 0.7426 | 0.7404 |
| 0.6593        | 4.0   | 1768 | 0.8127          | 0.7265   | 0.7564    | 0.7265 | 0.7245 |
| 0.5014        | 5.0   | 2210 | 0.8670          | 0.7507   | 0.7631    | 0.7507 | 0.7436 |
| 0.48          | 6.0   | 2652 | 0.7473          | 0.7694   | 0.7739    | 0.7694 | 0.7623 |
| 0.3505        | 7.0   | 3094 | 0.7647          | 0.8016   | 0.8039    | 0.8016 | 0.7991 |
| 0.3223        | 8.0   | 3536 | 0.7499          | 0.8016   | 0.8015    | 0.8016 | 0.7990 |


### Framework versions

- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1