htmlsift model weights
Weights for htmlsift, a web main-content extractor (HTML in, clean text / HTML / markdown out).
mini/— int8 ONNX bundle, runs on CPU.pip install htmlsiftbase/— 311M encoder bundle (Granite r2, first 10 layers), GPU-capable.pip install htmlsift[full]
The weights are fetched automatically on first use — you don't download them by hand.
License and attribution
Weights are licensed Apache-2.0. They are trained on WebMainBench (Apache-2.0) — please attribute WebMainBench when you use them.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support