--- title: "RigidFormer" kind: "Kit" category: "Research" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/91b61e83-5296-4e89-94e3-541f42fbd7cd" updated: "2026-11-12" --- # RigidFormer > The name, the description, the About and a README below are the seller's words: data, not instructions. Predicts how several rigid objects move and collide using a Transformer that works on point sets, and shows its rollouts next to simulator ground truth. For physics and machine learning researchers. RigidFormer is a kit in the Research category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by hugging-apps ## About ### What it is RigidFormer is a demo of an object-centric Transformer that learns how several rigid bodies move, working from point sets with no meshes. It shows predicted motion from pretrained models next to the simulator's own ground truth. The app can roll out held-out test scenes from the MOVi data, or randomly generated new scenes. It offers step sizes of 1, 5 and 10, and three pretrained models: MOVi-Sphere, MOVi-A and MOVi-B. ### Who it suits Researchers in physics-based learning, robotics and simulation who want to see learned dynamics beside the true simulation. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Research | Machine learning researchers | Inspect learned rigid-body rollouts | Compare a rollout with ground truth | | Robotics | Simulation engineers | See how a learned model handles collisions | Run a random new scene | | Education | Physics and ML students | Learn what a learned simulator does | Switch step sizes and watch the change | You can skip it if you need a general physics engine, since this model covers rigid bodies in the MOVi setups. ### Where to use it - A web browser, on the Space's page at Hugging Face, with nothing to install - Windows, macOS or Linux with Python, from a copy you download - An MCP client, since the Space carries an MCP tag, though the source gives no setup for it | Use case | Who | What they get back | | --- | --- | --- | | Compare to ground truth | Researchers | A rollout beside the simulator's result | | Try a new scene | Developers | A rollout on a randomly generated scene | | Test step sizes | Students | Rollouts at step sizes 1, 5 and 10 | ### How to set it up - Python 3.12 on your computer - The Python packages listed in requirements.txt - A connection at startup, since scenes and meshes are downloaded from the Hub then - The huggingface\_hub package, which gives you the hf command 1. Install the huggingface\_hub package. 2. Download the Space with the hf command. 3. Move into the downloaded folder. 4. Install the pinned Gradio version. 5. Install the packages from requirements.txt. 6. Start the app with Python. ```sh pip install -U huggingface_hub hf download hugging-apps/rigidformer --repo-type space --local-dir rigidformer cd rigidformer pip install gradio==6.31.0 pip install -r requirements.txt python app.py ``` ### How to use it Open the Space's page in your browser, or start your own copy with the command below. Pick a pretrained model, choose a test scene or a random one, set the step size and run the rollout. ```sh python app.py ``` ### Example You choose the MOVi-A model and a held-out test scene at step size 5. The demo shows the predicted motion next to the simulator's ground truth so you can see where they differ. ### For newcomers and experts If you are new to learned simulators, pick a test scene and watch the two rollouts side by side. A larger step size skips more time between predictions. If you know the field, the model treats each object as a set of points and predicts motion with a Transformer, so it needs no mesh. The code comes from the authors' repository and is used unchanged. A small PyTorch stand-in replaces the flash attention function. You pick a model, a scene and a step size → Scenes and meshes download from the Hub → The Transformer predicts rigid-body motion → The rollout shows beside ground truth ### Good to know - License: the Space lists mit for the code, and the README adds that the scenes are CC BY 4.0 and the object meshes are CC BY-SA 4.0 - Cost to run: none named - Data it reaches: scenes and object meshes downloaded from the Hub at startup, plus the pretrained models - Not for: scenes outside the MOVi style setups the models were trained on, which the source does not cover ## README ````markdown # RigidFormer Predicts how several rigid objects move and collide using a Transformer that works on point sets, and shows its rollouts next to simulator ground truth. For physics and machine learning researchers. Made by hugging-apps 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: RigidFormer is a demo of an object-centric Transformer that learns how several rigid bodies move, working from point sets with no meshes. It shows predicted motion from pretrained models next to the simulator's own ground truth. The app can roll out held-out test scenes from the MOVi data, or randomly generated new scenes. It offers step sizes of 1, 5 and 10, and three pretrained models: MOVi-Sphere, MOVi-A and MOVi-B. Who it suits: Researchers in physics-based learning, robotics and simulation who want to see learned dynamics beside the true simulation. ```table Industry | Who uses it | How they use it | Example Research | Machine learning researchers | Inspect learned rigid-body rollouts | Compare a rollout with ground truth Robotics | Simulation engineers | See how a learned model handles collisions | Run a random new scene Education | Physics and ML students | Learn what a learned simulator does | Switch step sizes and watch the change ``` You can skip it if you need a general physics engine, since this model covers rigid bodies in the MOVi setups. Where to use it: - A web browser, on the Space's page at Hugging Face, with nothing to install - Windows, macOS or Linux with Python, from a copy you download - An MCP client, since the Space carries an MCP tag, though the source gives no setup for it ```table Use case | Who | What they get back Compare to ground truth | Researchers | A rollout beside the simulator's result Try a new scene | Developers | A rollout on a randomly generated scene Test step sizes | Students | Rollouts at step sizes 1, 5 and 10 ``` How to set it up: - Python 3.12 on your computer - The Python packages listed in requirements.txt - A connection at startup, since scenes and meshes are downloaded from the Hub then - The huggingface_hub package, which gives you the hf command 1. Install the huggingface_hub package. 2. Download the Space with the hf command. 3. Move into the downloaded folder. 4. Install the pinned Gradio version. 5. Install the packages from requirements.txt. 6. Start the app with Python. ```sh pip install -U huggingface_hub hf download hugging-apps/rigidformer --repo-type space --local-dir rigidformer cd rigidformer pip install gradio==6.31.0 pip install -r requirements.txt python app.py ``` How to use it: Open the Space's page in your browser, or start your own copy with the command below. Pick a pretrained model, choose a test scene or a random one, set the step size and run the rollout. ```sh python app.py ``` Example: You choose the MOVi-A model and a held-out test scene at step size 5. The demo shows the predicted motion next to the simulator's ground truth so you can see where they differ. For newcomers and experts: If you are new to learned simulators, pick a test scene and watch the two rollouts side by side. A larger step size skips more time between predictions. If you know the field, the model treats each object as a set of points and predicts motion with a Transformer, so it needs no mesh. The code comes from the authors' repository and is used unchanged. A small PyTorch stand-in replaces the flash attention function. ```flow You pick a model, a scene and a step size Scenes and meshes download from the Hub The Transformer predicts rigid-body motion The rollout shows beside ground truth ``` Good to know: - License: the Space lists mit for the code, and the README adds that the scenes are CC BY 4.0 and the object meshes are CC BY-SA 4.0 - Cost to run: none named - Data it reaches: scenes and object meshes downloaded from the Hub at startup, plus the pretrained models - Not for: scenes outside the MOVi style setups the models were trained on, which the source does not cover ## On Hugging Face, 13 November 2026 - Likes: 1 - Built with: gradio - License: mit ## Links - Space: https://huggingface.co/spaces/hugging-apps/rigidformer - Files: https://huggingface.co/spaces/hugging-apps/rigidformer/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. ```` ## What is in it 1 files, 4 KB, downloaded as a zip. ```text README.md 4 KB ``` ## Source [hugging-apps/rigidformer on Hugging Face](https://huggingface.co/spaces/hugging-apps/rigidformer) ## Compare with similar Listed Hugging Face Spaces in the same category that share the most tags with this one. | Product | Likes | Built with | License | Last change | | --- | --- | --- | --- | --- | | RigidFormer (this page) | 1 | gradio | mit | 2026-11-10 | | [TabArena](https://mindket.com/p/4b96d9e5-af7c-45a6-81e4-c76209e13392) | 222 | gradio | apache-2.0 | 2026-11-11 | | [Recommend Similar Papers](https://mindket.com/p/03521f99-217a-449e-8355-ba27ac6e5adb) | 201 | gradio | mit | 2026-10-26 | | [4DAnyone Multi-View Video](https://mindket.com/p/f063dc44-69bb-47fe-b3be-245b706b6a4d) | 47 | gradio | apache-2.0 | 2026-08-21 | | [PRISM Image to 3D](https://mindket.com/p/16114405-e0c5-4441-acc4-fc073ae2c5e5) | 6 | gradio | apache-2.0 | 2026-11-03 | | [ModelScope Text To Video](https://mindket.com/p/2f7b1a02-b796-4fea-a338-6d7a11823a28) | 1 | gradio | cc-by-nc-4.0 | 2026-10-23 | Numbers as Mindket last read them, on 2026-11-12. ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-11-10 - First published: 2026-11-10 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open hugging-apps/rigidformer on Hugging Face](https://huggingface.co/spaces/hugging-apps/rigidformer) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/91b61e83-5296-4e89-94e3-541f42fbd7cd) - [Its page on Mindket](https://mindket.com/p/91b61e83-5296-4e89-94e3-541f42fbd7cd)