Deploying with Modelbit

Modelbit deployments each represent a collection of Python source code, model artifacts and environment dependencies. And each deployment has a REST API, logs showing recent calls and errors, and related settings around monitoring and resources.

For more information around creating a deployment, check out:

• Custom Python environments for specifying the pip and apt-get packages to install
• Private packages for including your private .whl packages
• Compute environments to use GPUs

After you've deployed, you'll want to call your model:

• Single inference requests shows the syntax typical for use cases where you want a single inference from your model.
• Batch inference requests shows the typical syntax for sending and receiving many inferences at once.
• DataFrame mode is for deployments that expect to receive pandas DataFrames.

And you can monitor your deployments with:

• Slack alerts to get notified when there are errors
• Logs integrations to stream logs into Datadog and Snowflake