SmolVLA is a compact vision-language-action policy for robot control from
visual observations, robot state, and language instructions. It targets practical fine-tuning and edge deployment use
cases.
This model was trained and exported with Physical AI Studio for local or Hugging Face-hosted robot inference.
Model Details
Policy: smolvla
Runtime library:physicalai
Generated by: Physical AI Studio
Intended Use
Use this model for language-conditioned robot inference in setups matching the training dataset, robot embodiment,
camera viewpoints, and task phrasing. Validate behavior in simulation or a safe test cell before running on hardware.
Dataset
This model was trained from the Physical AI Studio dataset named Dice-cleanup-2.
Model Package
Load the model from the root directory when possible. The root manifest.json is the package entry point, and
backend-specific manifests live under exports/<backend>/manifest.json.
Backend
Artifact
Intended Use
torch
exports/torch/smolvla.pt
Canonical checkpoint and Python inference
Training Environment
Environment: Empty
yaml
1name: Empty
2robots:[]3cameras:[]
I/O Specification
torch
Inputs
Name
Type
Shape
Dtype
state
STATE
[6]
float32
images.gripper
VISUAL
[3, 480, 640]
float32
images.overview
VISUAL
[3, 480, 640]
float32
task
LANGUAGE
[48]
string
Outputs
Name
Type
Shape
Dtype
action
ACTION
[50, 6]
float32
Running Inference
Installation
uv pip install physicalai numpy
The following smoke test verifies that the package loads and accepts tensors with the declared shapes. Replace the dummy
values with observations from your robot runtime before using the model for control.
Set MODEL_PATH to this local model directory or to the Hugging Face repository id after upload.
Running A Robot Control Loop
For a blocking control loop similar to PhysicalAI's examples/runtime/sync_inference.py, start from the training robot
and camera names exported above. Local device handles are placeholders because ports, camera paths, and stream URLs are
not included in published model metadata.
bash
1python examples/runtime/sync_inference.py \2 --robot so101 \3 --port /dev/ttyACM0 \4 --calibration ./calibration.json \5 --model path/to/model \6 --camera overhead:uvc:/dev/video0 \7 --task "Move the dice into the cup"\8 --device CPU
Training / Reproducing Training
Import this model in Physical AI Studio and start a new training job using it as the base model. Studio will preserve
the training lineage through the parent model relationship.
To reproduce behavior on your own hardware, match the exported I/O specification, robot type, camera viewpoints,
control frequency, language prompts, and calibration values from environment.json as closely as possible.
Evaluation
No task-specific evaluation metrics were exported with this generated card. Add validation results, success rates, and
hardware test conditions before publishing externally.
Limitations And Safety
Robot policies can behave unpredictably outside their training distribution. Validate camera viewpoints, lighting,
object placement, calibration values, robot embodiment, and task wording before autonomous operation. Use hardware
limits, emergency stops, supervision, and staged validation.
