Installation

Download the latest version by running:

pip install deeploy

Connecting to Deeploy

In order to initialize the Python client you need a personal key pair (account level access) or Deployment API token (deployment level access), the domain of your Deeploy installation and a workspace id. This ID can be found in the Workspace settings or the details of Deployments in that Workspace.

Use the following code to initialise the Deeploy Client:

from deeploy import Client

client = Client(
  # if using the Deeploy cloud version 'host' is 'app.deeploy.ml'
  host="example.deeploy.ml",
  workspace_id="b6d8c781-2526-4e03-9b43-4c1a62d064db",
  access_key="DPAexample",
  secret_key="Sexample",
)

Model and explainer Frameworks

Deeploy supports the following model frameworks with pre-build model and explainer images to make model deployments easy:

• Models
  • Scikit-Learn
  • XGBoost
  • LightGBM
  • PyTorch
  • Tensorflow
• Explainers
  • Anchors
  • Shap Kernel

For more information see here

Next to prebuild images Deeploy also supports custom model and explainer images. For more information about custom Docker Deployments we recommend to start here. You can find an example in the create Deployment section of the documentation.

Up next

The Python client currently includes the following functionalities:

• Create and update Deployments
• Inference Deployments
• Retrieve prediction logs
• Evaluate predictions
• Submit actuals for predictions
• Generate metadata.json
• Generate reference.json

For more information about authentication check this section