Verta
Search…
Quickstart
Get started with Verta in 5 minutes.

1. Obtain your Verta Credentials

  • Sign-up on the Verta instance setup for your organization. If you are using our trial instance, you can sign up here
  • Log into the Verta Web App and visit the Profile page to find your developer key.
Note that your developer key is unique to you. As with a password, don't share it with others!

2. Setup the Verta Client

The Verta client supports Python 2.7 & 3.5–3.8!
1
# via pip
2
python -m pip install verta
Copied!
Python 2.7 & 3.6–3.8 is supported through conda, as well.
1
# via conda
2
conda install verta -c conda-forge
Copied!
Following Python best practices, we recommended creating a virtual environment using venv or conda.

3. Integrate the Verta package into your workflow

1. Create a Verta client object to connect to the Verta server.

1
from verta import Client
2
client = Client(HOST, email=VERTA_EMAIL, dev_key=VERTA_DEV_KEY)
Copied!

2. Version your models

1
proj = client.set_project("Fraud Detection")
2
expt = client.set_experiment("Recurrent Neural Net")
Copied!
1
run = client.set_experiment_run("Two-Layer Dropout LSTM")
2
3
run.log_hyperparameter("num_layers", 2)
4
run.log_hyperparameter("hidden_size", 512)
5
run.log_hyperparameter("dropout", 0.5)
Copied!

3. Associate metadata with your models

1
run.log_metric("accuracy", 0.95)
2
run.log_tags(["experiment1"])
Copied!

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
Last modified 4mo ago