DistilHuBERT-Finetuned GTZAN Open-Source Audio Classification Model - Accurately Classify Music with an Accuracy of 86%

Distilhubert Finetuned Gtzan

Developed by f0ghedgeh0g

Audio classification model fine-tuned on the GTZAN music classification dataset based on the DistilHuBERT architecture, achieving 86% accuracy

Audio Classification

Transformers

Open Source License:Apache-2.0 #Music Classification #Audio Feature Extraction #Lightweight Model

Downloads 39

Release Time : 1/2/2025

Model Overview

This model is a lightweight audio processing model optimized for music genre classification tasks, capable of recognizing 10 different music genres

Model Features

Efficient Music Classification

Optimized for the GTZAN music dataset, accurately identifying 10 music genres

Lightweight Architecture

Based on the distilled DistilHuBERT architecture, reducing computational resource requirements while maintaining performance

High Accuracy

Achieves 86% classification accuracy on the GTZAN test set

Model Capabilities

Music Genre Classification

Audio Feature Extraction

Music Content Analysis

Use Cases

Music Industry

Music Streaming Auto-Tagging

Automatically tagging uploaded music with genre categories for music platforms

86% accuracy in automatic classification

Music Recommendation System

Improving recommendation algorithms based on music content similarity analysis

Academic Research

Music Information Retrieval

Supporting musicology research and music content analysis

🚀 distilhubert-finetuned-hyperparam-gtzan

This model is a fine - tuned version of ntu-spml/distilhubert on the GTZAN dataset. It offers a practical solution for audio classification tasks, achieving notable accuracy on the evaluation set.

🚀 Quick Start

This model is a fine-tuned version of ntu-spml/distilhubert on the GTZAN dataset. It achieves the following results on the evaluation set:

Loss: 1.2045
Accuracy: 0.86

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

🔧 Technical Details

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.7469	1.0	113	1.3737	0.57
0.7973	2.0	226	1.5247	0.57
0.6831	3.0	339	0.8961	0.74
0.4573	4.0	452	0.8638	0.76
0.1874	5.0	565	0.7839	0.81
0.0829	6.0	678	1.0174	0.79
0.0306	7.0	791	0.9393	0.81
0.004	8.0	904	0.9737	0.85
0.1209	9.0	1017	1.0625	0.8
0.0237	10.0	1130	1.3653	0.8
0.0164	11.0	1243	1.3065	0.81
0.0007	12.0	1356	1.1272	0.83
0.0004	13.0	1469	1.3226	0.83
0.0001	14.0	1582	1.6092	0.82
0.0001	15.0	1695	1.2045	0.86
0.0002	16.0	1808	1.1312	0.85
0.0003	17.0	1921	1.0911	0.86
0.0	18.0	2034	1.1983	0.84
0.0001	19.0	2147	1.1363	0.85
0.0	20.0	2260	1.2547	0.85
0.0002	21.0	2373	1.2394	0.84
0.0001	22.0	2486	1.5508	0.85
0.0	23.0	2599	1.2689	0.83
0.0	24.0	2712	1.2343	0.83
0.0003	25.0	2825	1.2313	0.81
0.0	26.0	2938	1.2217	0.83
0.0	27.0	3051	1.1596	0.84
0.0	28.0	3164	1.1081	0.85
0.0001	29.0	3277	1.1394	0.85
0.0	30.0	3390	1.1215	0.85

Framework versions

Transformers 4.48.0.dev0
Pytorch 2.6.0+cu126
Datasets 3.2.0
Tokenizers 0.21.0

📄 License

This project is licensed under the Apache-2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご