Model Asset Import Settings
The Model Asset Import Settings window lets you do the following:
- Change settings that affect how Sentis imports a model.
- Check the properties of an imported and optimized model.
To access this window, select an imported model asset in the Project window.
General settings
| Property | Description |
|---|---|
| Open | Open the model in the default application for ONNX files. Refer to Open a model as a graph for more information. |
| Optimize Model | When enabled, Sentis optimizes the model when you import it. The default is enabled. |
| Round Denormal Weights to 0 | When enabled, Sentis rounds very small float values (denormals) to 0. If you disable this setting, values in a model might be less precise but the model might run more slowly. The default is enabled. |
| Apply | Save changes you've made to the settings. |
| Revert | Revert changes you've made to the settings. |
Imported Object settings
Metadata
This section only appears if the model file contains ONNX metadata.
Inputs
This section shows the total number of inputs the model has, and a list of the inputs.
| Property | Description |
|---|---|
| name | The name of the input. |
| shape | The tensor shape of the input. For example (1, 1, 28, 28) means the model accepts a tensor of size 1 × 1 × 28 × 28 exactly. If a dimension in the shape is a string instead of a number, for example batch_size or height, the dimension can be any size (a dynamic input). Refer to Model inputs for more information about dynamic inputs. |
| dataType | The data type of the input. The possible values are float and int. |
Outputs
This section shows the total number of outputs the model has, and a list of the outputs.
| Property | Description |
|---|---|
| name | The name of the output. |
| shape | The tensor shape of the output. Sentis tries to precalculate this shape from the model. If a dimension in the shape is a question mark (?), Sentis can't calculate the size of the dimension, or the size depends on the input (a dynamic output). If the entire tensor shape is Unknown, Sentis can't calculate the number of dimensions, or the number of dimensions is dynamic. |
Layers
This section shows the total number of layers the model has, and a list of the layers in the order Sentis runs them.
| Property | Description |
|---|---|
| Type | The type of layer. Refer to Supported ONNX operators for more information. |
| name | The name of the layer. |
| inputs | The names of the inputs to the layer. Possible values are a model input from the Inputs section, another layer, or a constant. |
| Properties | The properties of the layer. The properties will depend on the type of layer. Refer to Supported ONNX operators for more information. |
Constants
This section shows the total number of constants the model has and the total number of weights, and a list of the constants.
| Property | Description |
|---|---|
| Type | The type is always Constant. |
| name | The name of the constant. |
| weights | The tensor shape of the constant. If the tensor shape is empty - () - the constant is a scalar (a zero-dimensional tensor). |
Model information
This section shows the total size of the model, its source, the ONNX opset version, and the tool that was used to create the model file.
Errors
This section only appears if there was a problem when Sentis imported and optimized the model, and the problems mean you can't run the model.
Warnings
This section only appears if there was a problem when Sentis imported and optimized the model. You can usually still run the model, but you might get unexpected results.
You can add a custom layer to implement a missing operator. Refer to the CustomLayer example in the sample scripts for a working example.