This template is made for using the Pytorch framework with Hydra for the configuration file and WandB for logging information. The template is designed to work with the CSV file from the Iris dataset available on Kaggle: Iris Dataset.

To make it work directly, you need to download the CSV file and place it in a directory named data. You should have the following structure:

Directory: D:\code\TemplatePytorchProject\data

Mode                 LastWriteTime         Length Name
----                 -------------         ------ ----
-a----         9/11/2024  11:39 AM           4551 iris.data.csv

Ensure the file is named iris.data.csv or modify the script accordingly to match the name.

Additionally, configuration parameters (such as dataset path, training parameters, etc.) are managed using Hydra. You can modify these settings by editing the configuration files located in the conf directory.

This template is based on the project: TimeSerieClassifierOnMusic.

The main directories and files in the project are structured as follows:

TemplatePytorchProject
│
├── conf
│   ├── data
│   ├── models
│   ├── predict
│   ├── trainer
│   ├── wandb
│   ├── default.yaml
│   ├── template.py
│   └── __init__.py
├── data
├── model
├── src
│   ├── data
│   │   ├── data.py
│   ├── models
│   │   ├── Model.py
│   ├── predict
│   │   ├── predict.py
│   ├── trainer
│   │   ├── train.py
├── sweep_file
│   ├── wandb_sweep.yaml
├── main.py
├── predict.py
└─── README.md

1. Create a virtual environment:

   python -m venv dev_env

2. Activate the virtual environment:

   ./dev_env/Scripts/Activate

3. Upgrade pip and install necessary packages:

   python.exe -m pip install --upgrade pip
   pip install -U pandas
   pip install -U numpy
   pip install -U wandb
   pip install -U hydra-core
   pip install -U torch --index-url https://download.pytorch.org/whl/cu121
   pip install -U torchaudio --index-url https://download.pytorch.org/whl/cu121
   pip install -U torchvision --index-url https://download.pytorch.org/whl/cu121
   pip install -U torchinfo
   pip install -U scikit-learn
   pip install -U matplotlib
   pip install -U tqdm
   

If you install new packages, update the requirements.txt file:

pip freeze > requirements.txt

Run the main.py script to train the model:

python main.py

Use wandb sweep with sweep_file\wandb_sweep.yaml file to make the exploration of hyper-parameter of the model:

wandb sweep .\sweep_file\wandb_sweep.yaml

this command will give you the command line to launch a agent. It's possible to launch multiple agents at once.

Use predict.py for using your save model:

python predict.py

Detailed configuration files are located in the conf directory. They are organized as follows:

• conf/data
• conf/models
• conf/predict
• conf/trainer
• conf/wandb

Refer to the respective YAML files for parameter settings and modify them as needed for your experiments.

We use Weights & Biases for experiment tracking. Ensure you have your API key configured in conf/wandb/default.yaml.