Skip to content

B. Learn Basics

Understanding the Application Folder

In the downloaded Azimuth folder from step A, the following structure exists:

azimuth  # Root directory
├── azimuth
   └── # Back End # (7)
├── azimuth_shr
   └── # User dataset, models, and code # (1)
├── config
   └── # User configs # (2)
├── webapp
   └── # Front End # (8)
├── .app_env  # (3)
├── docker-compose.yml  # (4)
├── Makefile  # (5)
└── README.md # (6)
  1. Where to put your data, model and code, if relevant. azimuth_shr stands for azimuth shared, because it contains user artifacts that are shared with the application. Azimuth provides default artifacts already to load common dataset and models.
  2. Where to put all configs. Example configs are provided.
  3. Default values for env vars.
  4. Where the config and Docker images are specified.
  5. Available commands to use Azimuth.
  6. Instructions to launch the application.
  7. Only available when cloning the repo.
  8. Only available when cloning the repo.

Where to put your data, code and configs?

config and azimuth_shr are two folders where you will put different artifacts before you can run Azimuth on your dataset and models. They get mounted automatically on the Docker image.

The Config

The Azimuth config contains all the information to launch and customize Azimuth. It specifies which dataset and pipelines to load in the app, as well as other variables that allow for customization of the app. Most fields have default values and don't need to be explicitly defined in each config. The Configuration reference details all available fields.

When launching Azimuth, you may give the launcher a JSON config file. If you don't, the UI will simply prompt you to configure Azimuth from the config UI.

Different configuration examples are provided in the repo under config/examples, leveraging pretrained models from HuggingFace. The next step, C. Run on Your Use Case, will detail how to adapt an existing config to create your own.

Clearing the Cache

Azimuth keeps all artifacts in caching folders so that if you close the app and re-launch, it will load quickly. Once you are done with your analysis, you can delete the cache by running:

make clean

Run Our Demo to Verify Your Setup

Out-of-the-box, Azimuth can run on different demo data and models from HuggingFace (HF). Verify that your setup is working correctly by running a demo.

  1. In the terminal, from the azimuth folder (the root directory), execute the following commands. The first one installs the Google Drive downloading library. The second command downloads from Google Drive the demo data and model. Our demo is using a subset of the clinc_oos dataset from HuggingFace, with only 16 classes.

    pip install gdown
    make download_demo
    

    You cannot install gdown?

    Look at the following Discussion to download the data manually.

  2. Run our dummy or full demo (option a. or b.), based on how much time you have. If it is the first time that you are running the command, it will take additional time to download the Docker image (~15 min).

    1. If you don't have a lot of time and just want to verify your setup, you can run our dummy CLINC demo (~2min):
      make CFG_PATH=/config/development/clinc_dummy/conf.json launch
      
    2. If you have a bit more time, run our full CLINC demo (~10min):
      make CFG_PATH=/config/development/clinc/conf.json launch
      
  3. The app will be accessible at http://localhost:8080 after a few minutes of waiting. The screen will indicate that the start-up tasks have started. When it is completed, the application will be loaded.

  4. Skim the Key Concepts section to get a high-level understanding of some concepts used throughout the application. If you are unsure what each screen allows you to do, the User Guide section walks you through all the available interactions on each screen.

Successful demo

Now that the demo is working, you can adapt the config to make it work on your dataset and model. Proceed to C. Run on Your Use Case.