galaxyHackers

Project aims to find and analyse galaxy clusters using AI in microwave and IK range

Installating Dependencies (UNIX)

We assume Python 3.10+ is installed. Dependencies can be managed via Poetry or pip.

Set Up Virtual Environment

1. Navigate to the project directory:

   cd galaxyHackers

Installing Dependencies

Option 1: using Poetry 2.0.0+

1. Enter the Poetry environment:

   poetry env activate

2. Install the project dependencies:

   poetry install

Option 2: using pip

If Poetry doesn't suit your needs or it fails to fetch all required libraries, you can use pip to install all dependencies.

1. (Optional) Set up and activate the virtual environment:

   python3.10 -m venv venv
   source ./venv/bin/activate

2. Install the necessary packages using pip:

   pip install torch torchvision timm torch_optimizer tqdm
   pip install numpy pandas matplotlib scikit-learn Pillow
   pip install astropy astroquery pixell dynaconf wget
   pip install comet_ml h5py ultralytics mlflow

Training Models

Once dependencies are installed, you can start training the models.

Train a model by running the main.py script:

python3 -m galaxy.main --models MODEL_NAME --epochs NUM_EPOCH --data DATASET

Available flags:

Flag	Description	Default / Options
--models	Model(s) to train	Baseline, ResNet18, EfficientNet, DenseNet, SpinalNet_ResNet, SpinalNet_VGG, ViTL16, AlexNet_VGG, CNN_MLP, YOLO12n, YOLO12s, YOLO12m, YOLO12l, YOLO12x (default: all except YOLO12s & YOLO12l)
--epochs	Number of training epochs	Default: 5
--mm	Momentum (for SGD or RMSprop)	Default: 0.9
--optimizer	Optimizer to use	Adam, SGD, Rprop, NAdam, RAdam, AdamW, RMSprop, DiffGrad (default: AdamW)
--repoptimizer	Wrap optimizer using RepOptimizer	Optional (see paper Re-parameterizing Your Optimizers rather than Architectures)
--segment	Generate segmentation maps after training	Optional
--data	Dataset to use	WISE (W1/W2 bands) or ACT (90, 150, 220 GHz from LAMBDA) (default: WISE)
--seed	Random seed (integer number)	Default: 1
--comet	Enable Comet.ml tracking	Optional
--mlflow	Enable MLflow tracking	Optional

Important

Only one optimizer can be passed in command-line.

In case script fails to download any dataset, it will address Legacy Survey website to download data of the infrared bands W1, W2 (for Russia available with VPN only).

Learning rate is found automatically using LR range test, so it is not passed using command-line.

For more information see A disciplined approach to neural network hyper-parameters: Part 1 -- learning rate, batch size, momentum, and weight decay

Running Multiple Models Simultaneously

You can run several models simultaneously with the same optimizer and learning rate.

---

Example Commands

Single model training:

python3 -m galaxy.main --models YOLO --epochs 5 --data ACT --segment

Multiple models training:

python3 -m galaxy.main --models AlexNet_VGG ResNet18 --epochs 20 --data WISE --segment

python3 -m galaxy.main --models --epochs 20 --data WISE --segment

Logging

The script supports Comet.ml and MLflow for experiment tracking.

Comet.ml

1. Rename .example.secrets.toml to .secrets.toml
2. In .secrets.toml pass your Comet API to the varible COMET_API_KEY and the name of the workspace into COMET_WORKSPACE
3. Enable --comet flag

Important

The script will not work without renaming .example.secrets.toml to .secrets.toml- this way you pass empty key.

MLflow

1. Enable --mlflow flag

Full documentation can be viewed here.