Skip to content
Other

blueprint-ai-hq-main/glassnet-readable-vision-21

a fully interpretable image classifier

2 min readupdated

Glassnet-Readable-Vision-21 is a unique image classification model that completely eliminates "black box" hidden layers. Each of the 21 neurons in its middle layer is an explicit formula that you can calculate by hand straight from image pixels. Running purely on numpy and safetensors without PyTorch, the model hits 84.62% accuracy on the Fashion-MNIST dataset with just 15,977 parameters. It is a great fit for researchers and developers exploring model transparency and interpretability.

Key features

  • Full transparency: 21 middle-layer neurons defined by explicit formulas (brightness, symmetry, texture, edges).
  • Strong accuracy: 84.62% on Fashion-MNIST, matching a similarly sized baseline model.
  • No PyTorch required: loading and inference run using only numpy and safetensors.
  • Honesty audit: 20 out of 21 concepts passed validation confirming they match their assigned names.

What you can do

  • Explore the architecture: inspect the source code and concept formulas to see how transparent models work under the hood.
  • Run classification experiments: use the numpy pipeline to classify 28x28 grayscale clothing items from Fashion-MNIST.
  • Ablation and testing: zero out or shuffle individual concepts to measure how each one impacts the final prediction.

Example

Load the model and work directly with its weights via numpy and safetensors to classify images without heavy deep learning frameworks.

Read moreDetails
Benchmarks: 84.62% accuracy (held-out), 84.86% baseline, 9.98% accuracy with zeroed middle layer
Limitations: works with Fashion-MNIST (28x28 grayscale), does not generate text descriptions of images, and is not suited for standard photos
Details: 15,977 parameters stored in model.safetensors with configuration in config.json
Discuss Telegram 0 18

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.0 GB · ready files0.5 GB
  • BF16weights 0.0 GB0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # blueprint-ai-hq-main/glassnet-readable-vision-21
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download blueprint-ai-hq-main/glassnet-readable-vision-21 --local-dir ~/models/glassnet-readable-vision-21
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of glassnet-readable-vision-21 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

blueprint-ai-hq-main/glassnet-readable-vision-21
Author
blueprint-ai-hq-main
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments