3-export.md

Goal 3: Export the model

Congratulations on successfully training and fine-tuning your model! Now it's time to actually use it.

To use the model in the sample web application, you first need to export the model.

Application notes

How to export the model

1. In the Use tab, select the Export button
2. Select TensorFlow.js
3. Choose the /exports folder in the project as the destination folder

Exported model folder structure

There are a few key elements in the main export directory. Your model in json format (model.json), its weights as binary shard files (.bin), and signature.json which contains information about your Lobe project.

This is what the folder structure of your exported model should look like:

exports/
  <Project Name> TensorFlowJS/
    example/
      package.json
      tfjsExample.ts
      README.md
    groupN-shardXofY.bin
    labels.txt
    model.json
    signature.json

Move model into web application

Copy the contents from your exported model folder into apps/web-bootstrap/public/model folder in the sample web application.

Success criteria

Your team will have achieved this goal when the following success criteria are met:

• Your team has exported TensorFlow.js model
• The model files have been moved into the public/model directory of the sample web application

Resources

Your team might find these resources helpful:

• Lobe - Exporting your model

Tips

• While you don't need the example/ folder or the labels.txt file, moving these into the web application will not cause any problems.

Final result

After exporting the model and copying it into the web application, you can now move to the final step: run the web app and use your model!