1.7B Part-Level Physics Checkpoint

Given a full-object point cloud and a target-part point cloud, this checkpoint predicts structured part identity, semantic descriptions, and intrinsic physical properties.

Model details

Item Value
Internal task name physics
Required geometry Object point cloud + target-part point cloud
Training lineage 1.7B initialization โ†’ physical semantic alignment
Training data To be announced
Evaluation data To be announced
Tested Transformers version 4.51.0

Inputs and outputs

The recommended point-cloud format is .npz with aligned arrays:

point   float32 [N, 3]
color   uint8   [N, 3]
normal  float32 [N, 3]

Object and part point clouds are normalized together by the inference pipeline. The structured result contains:

  • part_identity: part name and motion type;
  • semantic_description: basic, functional, movement, and grasp descriptions;
  • physical_properties: material, density, Young's modulus, hardness, Poisson's ratio, friction, graspability, and affordance.

Installation

Set CODE_REPO_URL to the URL of the anonymous source code repository.

git clone "$CODE_REPO_URL" source
cd source
python -m pip install -e .
bash scripts/install_cuda_extensions.sh

The model uses a custom Transformers architecture. Install the accompanying source code before loading the checkpoint; a generic transformers.pipeline is not supported.

Inference

Replace /path/to/checkpoint with the local checkpoint directory.

CUDA_VISIBLE_DEVICES=0 python inference_batch_intrinsic_physics_part.py \
  --model_path /path/to/checkpoint \
  --object_pcd examples/object.npz \
  --part_pcd examples/part.npz \
  --output outputs/intrinsic_physics_part.json

The entry point also accepts a JSON list through --input_json. See the main project README for the batch schema, path resolution rules, and output record format. The final prediction is stored in result, while raw_response keeps the direct generated text.

Evaluation

python -m eval intrinsic_physics_part PREDICTIONS \
  --output intrinsic_physics_part_metrics.json

Benchmark results are pending.

Intended use and limitations

This checkpoint is intended for research on physical grounding of 3D object parts and for producing candidate simulation parameters that are subsequently validated. It is not a physics simulator, and its estimates should not be treated as measured material properties.

Performance may degrade for incomplete scans, sparse or noisy point clouds, unseen materials, ambiguous part decompositions, unusual scales, or objects far outside the training distribution. Validate predictions before using them in robotics, safety-critical systems, or real-world engineering workflows.

License

The model weights are released under the Creative Commons Attribution-NonCommercial 4.0 International license. Commercial use is not permitted under this license. The accompanying source code is licensed separately under Apache-2.0. See the included license for Qwen3 and Sonata attribution.

Downloads last month
-
Safetensors
Model size
2B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for anonymous-abc123/UniPhysGen-1.7B-Physics

Finetuned
Qwen/Qwen3-1.7B
Finetuned
(1,416)
this model