Converting the model

AI Model Deployer

Product
AI Model Deployer
Product Version
1.1
Language
en-US

In the Model converter page, you can add a trained model, set the conversion parameters, and convert the model for running on the target device.

Current position

Tool bar

General information for the trained model

Conversions for the currently-selected trained model. You can add a new conversion by clicking "Add new conversion".

Navigation bar

Trained model list. You can add a new conversion by clicking "Add trained model".

"Model conversion parameters" area

"Converted model" area

Prerequisite

  • You have a trained model on your PC.

Add a trained model

  1. Navigate to the "model converter" page by clicking .

  2. Click "Add trained model". The Imported trained model window displays.

  3. Click "..." in the input box below "Trained model".

  4. Browse to the folder in which you saved the trained model, then select the model.

  5. Click "OK" to confirm.

    Then you can see the model is added in the model list.

  6. You can add more models by repeating step 2 to 5. You can add multiple trained models in a project.

You can use the following tools on the top-right corner to edit the added model:

View the structure of the trained model

Delete the added model

Set the model conversion parameters

  1. Enter the required data area in the parameter set. The fields marked with "*" are required.

    Scroll down to check all the parameters.

  2. You can add a different parameter set by clicking "Add new conversion".

You can use the following tools when you set the model conversion parameters:

Reset model conversion parameters

Delete the current parameter set

Convert model

  • After you set the model conversion parameters, click the Convert button to trigger the conversion.

    Then you can see that the model you add is marked as ready in the trained model list. For information on the supported model, refer to Supported model framework .

 

Status of the conversions of the selected model

① Number of ready conversions

② Number of not converted conversions

③ Number of incompatible conversions

④ Number of incompatible non-converted conversions

 

Status of the added model ⑤

  • Ready: at least one conversion is ready for this model

  • Not converted: no conversion is ready for this model

  • Not compatible: this trained model in not compatible with the target device

Note:

Compatibility of the conversions and the model

If you changed the version of target device after the model conversion, the trained model or the conversion might not be compatible with the target device any more.

  • If the trained model is not compatible with the changed target device, the model will be marked as not compatible, and you cannot add conversion for this model any more.

  • If the trained model is compatible with changed target device, but the existing conversion(s) is/are not compatible with it. The existed conversion will be marked as not compatible, and you can add new conversion for the changed destination device.

You can use the following tools to edit the converted model:

Convert the model again

Delete the conversion

Check error log

Download the model