music genre classifcation

Intro

Use fma dataset for genre classification with different deep learning architecture like Vision Transformer (ViT) arXiv:2010.11929 and other CNN like resnet.

Data

see https://github.com/mdeff/fma

Code

1. torch_transformer_classification.ipynb : preprocess dataset downsample sample to 22050 Hz and save them in .wav create and train transformeur
2. need to clean others notebooks and merge them

Usage

look at https://medium.com/france-school-of-ai/installer-sa-premi%C3%A8re-machine-avec-gpu-dans-le-cloud-98798fdc4406 (FR) for google cloud computing to train transformer I used Tesla v100 (GCP) test with python 3.7 and linux

1. Create a Python 3.7 environment

   conda create -n nom_env pyton=3.7
   conda activate nom_env
   conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c conda-forge
   pip install -U pip
   conda install matplotlib scikit-learn numpy pandas scipy -y
   conda install ipykernel
   python -m ipykernel install --user --name=fastai
   

2. Clone the repository

   git clone https://github.com/Julien-pour/music_classifcation.git
   

3. Install dependencies and dl dataset

   sudo apt-get update
   sudo apt-get install -y p7zip-full
   cd /data
   sudo wget https://os.unil.cloud.switch.ch/fma/fma_small.zip -O fma_small.zip   
   #use python for unzip  (problem when using 7z: sudo 7z x fma_small.zip && rm fma_small.zip)
   sudo wget https://os.unil.cloud.switch.ch/fma/fma_metadata.zip wget -O fma_metadata.zip
   sudo 7z x fma_metadata.zip && rm fma_metadata.zip
   

4. To launch jupyter notebook for GCP

   jupyter notebook --no-browser --port=6969 --ip='0.0.0.0'
   

5. look for vram (in terminal)

   watch -n 0.2 nvidia-smi
   

To do