Supported models

You can import open-source models into your Sentis project. Explore the following sections to understand the models Sentis supports, and find an appropriate model for your project.

Pre-trained models

There are various sources to find pre-trained models, which may either be available in the ONNX format or in a format that you can convert. Examples include:

If you want to train your own models, refer to the following links:

Models from Hugging Face

You can access validated AI models for use with Sentis from Hugging Face. Models available from Hugging Face are already in the .sentis format to remove the need for ONNX conversion.

To find and models from Hugging Face, you can either:

  • Navigate to the Unity Hugging Face space and select a model under the Models section.
  • View models that are validated for Sentis, identified by the unity-sentis tag on the Models page.

To import and use the model in a Unity project, follow the instructions in the How to Use section on the model page .

ONNX models

You can download ONNX model files from the ONNX model zoo on GitHub. Sentis supports most ONNX model files with an opset version between 7 and 15. While versions below 7 or above 15 may still be imported into Sentis, the results obtained may be unexpected.

Unsupported models

Sentis doesn't support the following:

  • Models that use tensors with more than 8 dimensions
  • Sparse input tensors or constants
  • String tensors
  • Complex number tensors

Sentis also converts some tensor data types like bools to floats or ints. This might increase the memory your model uses.

Additional resources