Create and update Deployments
Preparation
In order to deploy models, you need a Personal Key Pair. For more information regarding authentication, see Authentication.
Another part of the preparation is getting familiar with the repository contract that Deeploy uses, you can find more information about that here. If your model is stored elsewhere and you do not yet adhere to the contract, you generate a reference.json.
Default
If you are not using a SageMaker integration or Azure Machine Learning integration, you can use our default Deployment methods.
Create a Deployment
See all Create Deployment arguments
from deeploy import CreateDeployment
from deeploy.enums import ModelType
create_options: CreateDeployment = {
"name": "Python Client",
"description": "Python Client Description",
"repository_id": "b6d8c781-2526-4e03-9b43-4c1a62d064db",
"branch_name": "master",
"commit": "b26af7790c1d3b27b953b37806b91cb6735c590b",
"model_type": ModelType.CUSTOM,
"model_serverless": True,
}
deployment = client.create_deployment(create_options)
Update a Deployment
See all Update Deployment arguments
from deeploy import UpdateDeployment
update_options: UpdateDeployment = {
"model_mem_request": 256,
"model_mem_limit": 256,
"model_cpu_request": 0.1,
"model_cpu_limit": 0.5,
}
updated_deployment = client.update_deployment(deployment_id, update_options)
SageMaker
Create a SageMaker Deployment
See all Create SageMaker Deployment arguments
from deeploy import CreateSageMakerDeployment
from deeploy.enums import ModelType
create_options: CreateSageMakerDeployment = {
"name": "Python Client",
"description": "Python Client Description",
"repository_id": "b6d8c781-2526-4e03-9b43-4c1a62d064db",
"branch_name": "master",
"commit": "b26af7790c1d3b27b953b37806b91cb6735c590b",
"model_type": ModelType.CUSTOM,
"model_instance_type": 'ml.c5.large',
}
deployment = client.create_sagemaker_deployment(create_options)
Update a SageMaker Deployment
See all Update SageMaker Deployment arguments
from deeploy import UpdateSageMakerDeployment
update_options: UpdateSageMakerDeployment = {
"model_instance_type": 'ml.c4.large',
}
updated_deployment = client.update_sagemaker_deployment(deployment_id, update_options)
Azure Machine Learning
Create an Azure Machine Learning Deployment
See all Create Azure Machine Learning Deployment arguments
from deeploy import CreateAzureMLDeployment
from deeploy.enums import ExplainerType, ModelType
create_options: CreateAzureMLDeployment = {
"name": "Python Client",
"description": "Python Client Description",
"repository_id": "b6d8c781-2526-4e03-9b43-4c1a62d064db",
"branch_name": "main",
"commit": "ea420035f347fad091e367a9ce2798aef6cae9eb",
"model_type": ModelType.CUSTOM,
"explainer_type": ExplainerType.CUSTOM,
"model_instance_type": "Standard_D2as_v4",
"model_instance_count": 1,
"explainer_instance_type": "Standard_D2as_v4",
"explainer_instance_count": 1,
}
azure_ml_deployment = client.create_azure_ml_deployment(create_options)
Update an Azure Machine Learning Deployment
See all Update Azure Machine Learning Deployment arguments
from deeploy import UpdateAzureMLDeployment
update_options: UpdateAzureMLDeployment = {
"model_instance_type": "Standard_D2as_v4",
"model_instance_count": 1,
"explainer_instance_type": "Standard_D2as_v4",
"explainer_instance_count": 1,
}
updated_azure_ml_deployment = client.update_azure_ml_deployment(deployment_id, update_options)
General
Update the Deployment description
from deeploy import UpdateDeploymentDescription
description_options: UpdateDeploymentDescription = {
"name": "Python Client v2",
"description": "Python Client Description v2",
}
updated_deployment = client.update_deployment_description(deployment_id, description_options)
Creating multiple Deployments from a monorepo
Deeploy supports creating multiple Deployments from the same repository using a contract path. The contract path is the path relative to the root of a repository where the deeploy contract for a Deployment is located. The idea is that you create your Deeploy contract path at a subfolder in the repository, as shown in the example below.
repository
│
├─ subfolder_1
│ ├─ model
│ │ └─ reference.json
│ ├─ explainer
│ │ └─ reference.json
│ ├─ transformer
│ │ └─ reference.json
│ └─ metadata.json
│
└─ subfolder_2
└─ model
└─ reference.json
To create a Deployment from a contract path, you provide the relative path from the repository root as input to the Python SDK or Core API. In the example above this is subfolder_1
or subfolder_2
.
from deeploy import CreateDeployment
create_options: CreateDeployment = {
# other options
"contract_path": "/subfolder_1",
}
deployment = client.create_deployment(create_options)
Check out the API reference for more information.