readme-for-group-A.md

Quick Guide for Pesto Model Training

This repo comes with MIR-1k downloaded inside data folder.

But in case you want to download the data by yourself, here's the tutorial:

On Mac:

1. Create a folder named data in project root dir: mkdir ./data
2. Download MIR-1k from kaggle and put it inside this folder.
3. Generate csv file:

   find MIR-1K/Vocals -name "*.wav" | sort > mir-1k.csv
   find MIR-1K/PitchLabel -name "*.pv" | sort > mir-1k_annot.csv

4. Start training!

   python src/train.py data=mir-1k logger=csv

   • Optionally, on mac you may add trainer=mps or trainer=cpu

On Windows

1. Create a folder named data in project root dir: mkdir ./data
2. Download MIR-1k from kaggle and put it inside this folder.
3. Create training data csv files:

   cd data
   Get-ChildItem -Recurse -Filter "Wavfile\*.wav" | Sort-Object -Property Name | Select-Object -ExpandProperty FullName > "..\mir-1k.csv"
   Get-ChildItem -Recurse -Filter "PitchLabel\*.pv" | Sort-Object -Property Name | Select-Object -ExpandProperty FullName > "..\mir-1k_annot.csv"   

4. Start training!

   python src/train.py data=mir-1k logger=csv