Deploying Arcee to your AWS VPC
Overview
The privacy of your data and models is paramount. Hence, you can deploy Arcee's entire software platform in your AWS VPC to keep your models where your data lives.
Configuring VPC
You can have Arcee create a new VPC or attach to one of your existing VPCs.
To configure your VPC for Arcee deployment, you'll need to set up CIDR blocks, public subnets, and private subnets. The example configuration provided demonstrates the necessary settings.
Here's the example configuration:
{
"environment": "arcee-in-vpc",
"domain_name": "myarcee.ai",
"zone_id": "ZONE",
"customer_account_id": "AWS ACCT ID",
"region": "AWS ACCT REGION",
"skip_aoss": true,
"skip_saml_bootstrap": true,
"platform_version": "vpc",
"arcee_dev": false,
"internal_lb": false,
"max_lambda_memory": 3007,
"vpc_config": {
"create": {
"cidr_block": "CIDR",
"public_subnets_cidr": ["PUBLIC CIDR"],
"private_subnets_cidr": ["PRIVATE SUBNETS"]
}
},
"cross_account_role_arn": "ARCEE CROSS ACCOUNT ROLE ARN",
"enable_quota_requests": false
}
The vpc_config section includes important details like the cidr_block, which defines the IP address range for your VPC.
The public_subnets_cidr array contains CIDR blocks for your public subnets.
The private_subnets_cidr array contains CIDR blocks for your private subnets.
These settings ensure that your VPC is configured to handle the deployment of Arcee models within your specified environment. Make sure to adjust the configuration values to match your specific requirements before deployment.
Required AWS Resources
Ensure you have an AWS VPC with configured subnets, a cross account role with necessary permissions, and a correctly set up security group before deploying Arcee.
Allow Egress to HuggingFace
If you do not allow egress to HuggingFace, you will be restricted in the variety of models you can utilize within Arcee
Provision F Subnet
To use AWS capacity blocks, you'll need to provision an F Subnet in your VPC. This step is necessary to ensure smooth operation and resource allocation.
Frequently Asked Questions
Arcee deployment refers to the process of setting up a Small Language Model (SLM) within your AWS Virtual Private Cloud (VPC) using the Arcee platform.
To troubleshoot deployment issues, check the Arcee logs and verify all configuration settings. Ensure that the deployment arguments, such as deployment name and model types, are correctly provided.
Yes, you can use an existing VPC to deploy Arcee. Ensure that the VPC meets the necessary requirements for running the Arcee stack.
You need permissions to create and manage AWS resources like EC2 instances and VPC configurations. Ensure your AWS account has the necessary roles and policies assigned.