WINTER real/bogus ML models, originally created by @aswinsuresh

This slimmed-down version uses pytorch rather than tensorflow.

pip install winterrb

• Clone the repository

git clone git@github.com:winter-telescope/winterrb.git

• Navigate to the repository

cd winterrb

• Create a conda environment with the required packages

conda create -n winterrb python=3.11

• Activate the environment

conda activate winterrb

• Install the package

pip install -e .

You need a data directory, containing a list of training classifications in csv format, named training_data.csv, and a data containing the corresponding avro alerts used for training. Specifically, you require a directory within the data directory named train_data containing the avro alerts. Each avro alert should be named with the format <id>.avro.

You can set the data directory using the bash environment variable WINTERRB_DATA_DIR.

export WINTERRB_DATA_DIR=/path/to/data

Then you can train a model using the notebook winterdrb_pytorch.ipynb.