NorviTech / AI models
Open models.
Your hardware.
Models and tools you can build into your own projects. Explore each release, then find its model card, files and documentation on Hugging Face.
NorviTech on Hugging FaceCompact CPU classifier
Armiger
Small enough for the edge.
Built for your media library.
A byte-level model that reads release titles and file trees to classify torrents and media. The compact ONNX INT8 release runs on a CPU and predicts eight classification heads in one pass.
- Runtime
- CPU · ONNX INT8
- Model size
- 8.3 MB
- Task
- Torrent & media classification
- Outputs
- 8 classification heads
- Licence
- Apache-2.0
One pass, multiple decisions
Predict content category, adult content, anime, audio language, title language, English subtitles, junk and owner scope together.
Start with the core
Download the model and calibration file for a compact deployment. The model card explains the outputs, intended uses and published evaluations.
Release details and evaluation results: Armiger model card.
Classification + swarm intelligence
Armiger Ultra
The complete toolkit.
One download.
The full-precision classifier, compact CPU fallback and modular plugins in one release. Add swarm health signals, release priors, catalog lookups, audio language checks, calibration and DHT tools to your pipeline.
- Runtime
- GPU / CPU · ONNX & PyTorch
- Full-precision model
- 32.2 MB
- CPU fallback
- Armiger INT8 included
- Toolkit
- 6 bundled plugins
- Licence
- Apache-2.0
Run the unified pipeline
The release includes the classifier, fallback model and an ArmigerUltraPipeline SDK. Its model card walks through setup and use.
Use the pieces you need
The six plugins are bundled with Ultra and published separately on Hugging Face for custom integrations. Explore their documentation from the model card.
Release details and evaluation results: Armiger Ultra model card.