BurnyCoder/bsf-dinov3-rabbits-readme-quickstart

Grassmannian BSF — README quickstart

This is a trained block-sparse featurizer reproduced from Goodfire's original repository at immutable commit 0bf2d9a6ae959452d57bc169374c8902135e0f02. It is part of the Block-Sparse Featurizers on DINOv3 Rabbits collection.

Exact recipe

• Source: README.md (Git blob d76be59a54cfc06203f1b56fc90361dd880e51ae)
• Constructor arguments: {"d": 768, "group_size": 3, "l0": 16, "n_groups": 256}
• bsf.train arguments: {"epochs": 60}
• Effective defaults: batch size 2048, SNR 0.1, learning rate 0.0004
• DINO backbone: facebook/dinov3-vitb16-pretrain-lvd1689m at 5931719e67bbdb9737e363e781fb0c67687896bc
• Reconstruction R²: 0.823179
• Mean active blocks: 16.0
• Dead groups: 0

The full machine-readable provenance, input hashes, environment, and metrics are in manifest.json.

Loading

The hardened loader and immutable catalog live in block-sparse-featurizer-experiments. Install that application from a revision containing this model's Hub commit:

git clone --recurse-submodules https://github.com/BurnyCoder/block-sparse-featurizer-experiments.git
cd block-sparse-featurizer-experiments
uv sync --frozen

from bsf_experiments.artifacts import restore_checkpoint
from bsf_experiments.hub_phase import (
    download_hub_checkpoint,
    get_hub_checkpoint_spec,
)
from bsf_experiments.types import PretrainedRecipe

spec = get_hub_checkpoint_spec(PretrainedRecipe.README_QUICKSTART)
path = download_hub_checkpoint(spec)
model, model_config = restore_checkpoint(path)

The application catalog pins a full Hub commit, file size, input width, and SHA-256 before this code restores the checkpoint. Loading avoids BSF retraining; new images still require the same DINOv3 patch-token extraction, positional-mean subtraction, and RMS scaling.

Limitations and licenses

The checkpoint was trained only on the bundled 300-image rabbit dataset and should not be assumed to generalize to other data. Training intentionally preserved the original unseeded stochastic behavior, so an independent reproduction need not be bit-identical.

Goodfire's BSF software terms are in LICENSE-goodfire.txt. DINOv3's license, which governs the backbone and its materials, is in LICENSE-dinov3.md; review those terms before redistribution or use.