CLI Commands and Supported Models

Commands List

By typing unleash in you terminal you can access to the supported CLI commands:

login

This command asks you to enter your credentials allowing fetch the information from your Unleash Live account. here is the sample output:

Note: If your credentials are incorrect, you would not see the

"Thank you for using Unleash Cloud CLI"

logout

By this command you can log out from your account.

list

You may use this command to list the AI model that you already have in your account. The output can be similar to the following, which shows the date, name and the description of the models:

Note: If you have not logged in already, this command will ask you to log in first and once the credentials are correct, it will show you the list of models.

clone

If you already have any model in your account, you can clone and modify your model. The CLI will give you the list of your model to select. You may navigate in the list using up and down arrows in the keyboard.

After selecting an AI model, the CLI asks you where to save the information. Note that the folder should be existed, otherwise the CLI will ask you the address again.

The  addonData.json file will be saved in the selected folder. This file contains the information regarding your model, and you can modify and push it back.

push

Once you have done your AI model modifications, you can use this command to push it back to your account. By running this command, the similar steps to init command will be executed. Note: If you have not changed the model weight file, you can skip the process of the upload weight.

init

By this command you can initialise a model and select it from the list of predefined models:

The only difference in deploying your model using above platforms is the file types that you need to provide. You may find the correct file types in each section accordingly, however, the whole process for a sample yolov3 model is explained in Yolov3 section.

Yolov3

By selecting the  Yolov3 model from the list, the CLI asks you to provide a path containing the related information for your model:

Notes:

  • The provided path should not have contained addonData.json file. Otherwise, the CLI gives warning and asks for the correct folder path.
  • The folder must contain the following files:
    1. The class names with the .names extension i.e. classes.names.
    2. The configuration file with the .cfg extension i.e. yolov3.cfg.
    3. The weight file with the .weights extension i.e. yolov3.weights.
    • if the folder does not contain any of the above files, it gives a warning.

Once the CLI examined the entered folder path and found the required files, it asks questions regarding the AI model information. Although each question has default value, but it is highly recommended filling in the questions with proper answers. You can change these fields in the process of command as well. Here is the sample answers to questions:

In this stage, you need to select the correct classes names, configuration file and weight file (if you have more than one file for each ion your folder). You may press  a to select all the listed files. Then the CLI uploads the files to your Unleash Live account and provides addonData.json file in your local folder.

Now, the model is created in your account you can access to it from the account:

You may modify the  addonData.json for variety of plugins.

Yolov5

The required file for Yolov5 model is only  .pt file. Please note that this Yolov5 model is based on the ultralytics implementation. You may follow the Yolov3 instruction for the rest of deployment process.


Open Neural Network Exchange (ONNX)

The required files for ONNX model are  .onnx and .names information. You may follow the Yolov3 instruction for the rest of deployment process.


Yolact

The required file for Yolov5 model is only  .pt file. You may follow the Yolov3 instruction for the rest of deployment process.

Detectron2

The required files for Detectron2 model are  .yml and .pth information. You may follow the Yolov3 instruction for the rest of deployment process.