Classify torrent metadata with a local ONNX model

Install the CPU model

Use Python 3.12 and Git in a terminal. This example uses Armiger v3.13.0 and its pinned v3.12.0 CPU model. Install using the GitHub URL below. The PyPI name armiger refers to a different project. The setup downloads the package and model files. Prediction runs locally.

python3 -m venv .venv
. .venv/bin/activate
git clone --depth 1 --branch v3.13.0 https://github.com/spencernorton/armiger.git
python -m pip install ./armiger
python -m armiger download --directory ./models-cpu --variant cpu

Classify a fictional movie release

Save this as classify_example.py in the directory where you downloaded the weights. The file names and size describe fictional metadata, so you do not need an actual torrent or video file.

from armiger import ArmigerStandaloneONNX

model = ArmigerStandaloneONNX(
    onnx_path='./models-cpu/armiger.int8.onnx',
    vocab_path='./models-cpu/vocab.json',
    calibration_path='./models-cpu/calibration.json',
)
result = model.predict({
    'name': 'Example.Movie.2001.1080p.WEB.x264',
    'files': [{'path': 'Example.Movie.2001.mkv', 'size': 1_000_000_000}],
})
print(result['category'])

Run python classify_example.py. This fixture prints movie. Treat the category as a prediction from the supplied name and file list. For a new feed, compare a sample of predictions with labels you trust before applying a filter. Armiger does not inspect media payloads or determine whether a download is safe.

Use the result in your own index

Keep the standalone model if your code already supplies metadata. For an existing Bitmagnet or Rats Search workflow, start with the Bitmagnet guide or Rats Search guide. Those setups use Armiger Ultra and explain the additional catalogue and swarm context. The pinned model card and files and evaluation notes describe model scope.

Back to Armiger