Weightprint

Weightprint

Measures what a fine-tune changed in a model by comparing its weights with its declared parent, in a static page. For people who want to check what a Hub model is made of.

Made by tardellirs on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Weightprint is a static web page that measures a model's weights and compares them with the parent named on its model card. It can tell a plain copy from a quantized model, a full fine-tune or an abliterated one. You can open it in a browser without installing anything.

Who it suits: It suits researchers, model users and Hub curators who want evidence of how a fine-tune differs from its parent, instead of trusting the model card alone.

Industry | Who uses it | How they use it | Example
AI research | Researchers | Measure how much a fine-tune moved the weights | Look at which layers changed against the parent
Model hosting | Hub curators | Compare what a card declares with what the weights show | Check a model that claims a certain parent
Software | Developers choosing a model | Check what a downloaded model is based on | Compare two models directly

If you only want to run models, not inspect them, you can skip it.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• A web browser on Windows, macOS or Linux, from a local copy of the page files

Use case | Who | What they get back
Browse the Wall | Anyone curious | One bar per measured fine-tune, lit where the weights moved
Look up a model | Researchers and users | The layers and parts that changed against its parent
Compare two models | Researchers | A direct comparison of any two models
Read the state of the Hub | Curators | Declared against measured across 120,976 models

How to set it up:

• A web browser, which is all the online page needs
• Python with pip, only if you want the files on your own computer

1. Open the Space's page on Hugging Face to use it online. Nothing needs installing.
2. For a local copy, install the Hugging Face hub package with pip.
3. Download the Space's files into a folder named weightprint.
4. Move into that folder and open index.html in a web browser.

pip install -U huggingface_hub
hf download tardellirs/weightprint --repo-type space --local-dir weightprint
cd weightprint

How to use it: Open the page and pick one of its three views.

Tool | What it does
Wall | A chart with one bar per measured fine-tune, brightened where its weights shifted
Any model | Type in a Hub id for a list of changed layers and parts compared with the parent, or compare any two models
State of the Hub | Numbers for 120,976 models: declared against measured, typical size of fine-tune changes, most-derived base models and exact byte copies

Example: In the Any model view, enter the id of a fine-tuned model from the Hub. The page shows which of its layers and parts differ from the parent that its card declares.

For newcomers and experts: Start with the Wall view and look at the lit bars. A bright stretch means the weights moved a lot there, and a dark one means they barely changed.

The page never downloads a model. To measure, it fetches each file's header and a slice of every tensor with range requests. Small tensors come in full, and larger ones give a 512-value sample. That keeps it light, but a window is a sample, so a change outside it can be missed. The full results sit in the tardellirs/weightprint dataset.

You enter a Hub model id
The page reads each file header
It reads part of every tensor
It compares them with the parent
The changed layers are shown

Good to know: The measurements come from samples of the tensors, so treat them as evidence and not as proof.

• License: Apache 2.0
• Cost to run: none named
• Data it reaches: model files on the Hub, read in part through range requests, and the tardellirs/weightprint dataset

On Hugging Face, 13 November 2026

• Likes: 6
• Built with: static
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/tardellirs/weightprint
• Files: https://huggingface.co/spaces/tardellirs/weightprint/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.