drowse browser: gemma3-1b-pt-candidate

This is a local browser build published as a candidate, not an install-ready signed-catalog release. The q4f32 build produced finite completions and repeatable capture/replay, but the matched-weight reference diverged after 11 generated tokens. Positional-encoding and long-context correctness remain unresolved. It must not be used as a numerically validated research runtime.

Download

Download this repository from Files and versions, or use:

hf download logitsml/drowse-web-gemma3-1b-pt-candidate --local-dir ./drowse-web-gemma3-1b-pt-candidate

These are browser-runtime artifacts, not a standard Transformers checkpoint. Model weights, tokenizer files, WebAssembly runtime and provenance are included. Browser installation still requires a compatible runtime and signed catalog; public file availability alone does not enable the app download button.

Provenance

Source: google/gemma-3-1b-pt. Exact file hashes and source revision are in UPLOAD-PROVENANCE.json. The build contains converted weights and a custom instrumented WebAssembly library; it is not the original checkpoint.

See LICENSES.md for component-specific licenses.

Gemma terms

Use, reproduction, modification and redistribution of these Gemma-derived artifacts are governed by the included Gemma Terms of Use, including Section 3.2 and the Gemma Prohibited Use Policy, incorporated here by reference. These use restrictions apply to all downstream use and distribution. See NOTICE.txt.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support