Quickstart

Get started with Verta in 5 minutes.

1. Setup the Verta Server

You have a few options here:

  • If you are on Verta Core or Verta Enterprise, you do not need to set up a Verta Server; please reach out to Verta Support for instructions for your server information.
  • You are only looking to run open-source ModelDB, head over to the ModelDB repo for installation instructions.

2. Setup the Verta Client

The Verta client supports Python 2.7 & 3.5–3.7!

# via pip
pip install verta

# via conda
conda install verta -c conda-forge

Following Python best practices, we recommended creating a virtual environment using venv or conda.

3. Obtain your Verta Credentials

  • On Verta Core or Verta Enterprise, log into the Verta Web App (e.g., https://app.verta.ai) and visit the Profile page to find your developer key.

    _images/web-app-profile.png

    Note that your developer key is unique to you. As with a password, don’t share it with others!

  • If using ModelDB open-source, you will not require any special credentials

4. Integrate the Verta package into your workflow

  1. Create a Verta client object to connect to the Verta server.
from verta import Client
client = Client(HOST, VERTA_EMAIL, VERTA_DEV_KEY)

If using open-source ModelDB, leave `VERTA_EMAIL` and `VERTA_DEV_KEY` blank.

  1. Version your models
proj = client.set_project("Fraud Detection")
expt = client.set_experiment("Recurrent Neural Net")

run = client.set_experiment_run("Two-Layer Dropout LSTM")

run.log_hyperparameter("num_layers", 2)
run.log_hyperparameter("hidden_size", 512)
run.log_hyperparameter("dropout", 0.5)
  1. Associate metadata with your models
run.log_metric("accuracy", 0.95)
run.log_tags(["experiment1"])

5. Check out your models!

Now that you have versioned a few models, you can interact with them in a variety of ways:

  • Build dashboards on the Verta Web App based on the models
  • Merge, branch, and manage all changes to your models
  • Share your models and reports with your organization or publicly
  • Deploy versioned models via Verta Deployment and Monitoring