Deployment

verta deployment create

Create an entity related to deployment.

For example, to create a new endpoint, run

verta deployment create endpoint “<endpoint path>”

verta deployment create [OPTIONS] COMMAND [ARGS]...

endpoint

Create a new deployment endpoint.

verta deployment create endpoint [OPTIONS] PATH

Options

-w, --workspace <workspace>

Workspace to use.

Arguments

PATH

Required argument

verta deployment download

Download resources relating to deployment.

For example, to download the Docker context for a model version, run

verta registry download dockercontext -m “<model version id>”

verta deployment download [OPTIONS] COMMAND [ARGS]...

dockercontext

Download a registered model version’s or experiment run’s context entry.

verta deployment download dockercontext [OPTIONS]

Options

-m, --model-version-id <model_version_id>

Model Version ID

-o, --output <output>

Required Filepath to write to

-r, --run-id <run_id>

Experiment Run ID

verta deployment get

Get detailed information about an entity related to deployment.

For example, to see information about an endpoint, run

verta deployment get endpoint “<endpoint path>”

verta deployment get [OPTIONS] COMMAND [ARGS]...

endpoint

Get detailed information about a deployment endpoint.

verta deployment get endpoint [OPTIONS] PATH

Options

-w, --workspace <workspace>

Workspace to use.

Arguments

PATH

Required argument

verta deployment list

List all instances of an entity related to deployment.

For example, to list all endpoints, run

verta deployment list endpoint

verta deployment list [OPTIONS] COMMAND [ARGS]...

endpoint

List all endpoints available.

verta deployment list endpoint [OPTIONS]

Options

-w, --workspace <workspace>

Workspace to use.

verta deployment predict

Making prediction to a deployment-related entity.

For example, to make a prediction to an endpoint, run

verta deployment predict endpoint “<endpoint path>” –data “<input data>”

verta deployment predict [OPTIONS] COMMAND [ARGS]...

endpoint

Making prediction via a deployed endpoint.

verta deployment predict endpoint [OPTIONS] PATH

Options

-d, --data <data>

Required Input for prediction. Must be a valid JSON string.

-w, --workspace <workspace>

Workspace to use.

Arguments

PATH

Required argument

verta deployment update

Update an entity related to deployment.

For example, to update an endpoint with a new model version, run

verta deployment update endpoint “<endpoint path>” -m “<model version id>” -s direct

verta deployment update [OPTIONS] COMMAND [ARGS]...

endpoint

Update an endpoint.

verta deployment update endpoint [OPTIONS] PATH

Options

--autoscaling <autoscaling>

Quantities for autoscaling. Must also provide –autoscaling-metric.

--autoscaling-metric <autoscaling_metric>

Metrics for autoscaling. Can only be used alongside –autoscaling.

-i, --canary-interval <canary_interval>

Rollout interval, in seconds. Can only be used alongside –strategy=canary.

-c, --canary-rule <canary_rule>

Rule to use for canary deployment. Can only be used alongside –strategy=canary.

--canary-step <canary_step>

Ratio of deployment to roll out per interval. Can only be used alongside –strategy=canary.

--env-vars <env_vars>

Environment variables to set for the model build. The format is –env-vars ‘{“VERTA_HOST”: “app.verta.ai”}’.

-m, --model-version-id <model_version_id>

Model Version to deploy. Cannot be used alongside –run-id.

-f, --filename <filename>

Path to JSON or YAML config file. Can’t be used alongside other options except for –workspace.

--resources <resources>

Resources allowed for the updated endpoint.

-r, --run-id <run_id>

Experiment Run to deploy. Cannot be used alongside –model-version-id.

-s, --strategy <strategy>

Strategy to use to roll out new deployment.

Options:

direct | canary

-w, --workspace <workspace>

Workspace to use.

Arguments

PATH

Required argument