azureml-examples/cli/deploy-moe-vnet-mlflow-lega...

111 строки
6.1 KiB
Bash

#!/bin/bash
set -e
# This is now a legacy script that uses the old way of creating a managed vnet endpoint. The new way is to use Workspace Managed VNet. See cli\deploy-managed-online-endpoint-workspacevnet.sh for the new way.
# This is the instructions for docs.User has to execute this from a test VM - that is why user cannot use defaults from their local setup
# <set_env_vars>
export SUBSCRIPTION="<YOUR_SUBSCRIPTION_ID>"
export RESOURCE_GROUP="<YOUR_RESOURCE_GROUP>"
export LOCATION="<LOCATION>"
# SUFFIX that was used when creating the workspace resources. Alternatively the resource names can be looked up from the resource group after the vnet setup script has completed.
export SUFFIX="<SUFFIX_USED_IN_SETUP>"
# SUFFIX used during the initial setup. Alternatively the resource names can be looked up from the resource group after the setup script has completed.
export WORKSPACE=mlw-$SUFFIX
export ACR_NAME=cr$SUFFIX
# provide a unique name for the endpoint
export ENDPOINT_NAME="<YOUR_ENDPOINT_NAME>"
# name of the image that will be built for this sample and pushed into acr - no need to change this
export IMAGE_NAME="mlflow"
# Yaml files that will be used to create endpoint and deployment. These are relative to azureml-examples/cli/ directory. Do not change these
export ENDPOINT_FILE_PATH="endpoints/online/managed/vnet/mlflow/endpoint.yml"
export DEPLOYMENT_FILE_PATH="endpoints/online/managed/vnet/mlflow/blue-deployment-vnet.yml"
export SAMPLE_REQUEST_PATH="endpoints/online/managed/vnet/mlflow/sample-request.json"
export ENV_DIR_PATH="endpoints/online/managed/vnet/mlflow/environment"
# </set_env_vars>
export SUFFIX="mevnet" # used during setup of secure vnet workspace: setup/setup-repo/azure-github.sh
export SUBSCRIPTION=$(az account show --query "id" -o tsv)
export RESOURCE_GROUP=$(az configure -l --query "[?name=='group'].value" -o tsv)
export LOCATION=$(az configure -l --query "[?name=='location'].value" -o tsv)
# remove all whitespace from location
export LOCATION="$(echo -e "${LOCATION}" | tr -d '[:space:]')"
export IDENTITY_NAME=uai$SUFFIX
export WORKSPACE=mlw-$SUFFIX
export ENDPOINT_NAME=$ENDPOINT_NAME
# VM name used during creation: endpoints/online/managed/vnet/setup_vm/vm-main.bicep
export VM_NAME="moevnet-mlflow-vm"
# VNET name and subnet name used during vnet worskapce setup: endpoints/online/managed/vnet/setup_ws/main.bicep
export VNET_NAME=vnet-$SUFFIX
export SUBNET_NAME="snet-scoring"
export ENDPOINT_NAME=endpt-vnet-mlflow-`echo $RANDOM`
# Get the current branch name of the azureml-examples. Useful in PR scenario. Since the sample code is cloned and executed from a VM, we need to pass the branch name when running az vm run-command
# If running from local machine, change it to your branch name
export GIT_BRANCH=$GITHUB_HEAD_REF
# need to set branch name manually if executed from main
if [ "$GIT_BRANCH" == "" ];
then
GIT_BRANCH="main"
fi
# We use a different workspace for managed vnet endpoints
az configure --defaults workspace=$WORKSPACE
export ACR_NAME=$(az ml workspace show -n $WORKSPACE --query container_registry -o tsv | cut -d'/' -f9-)
if [[ -z "$ACRNAME" ]]
then
export ACR_NAME=$(az acr list --query '[].{Name:name}' --output tsv)
fi
### setup VM & deploy/test ###
# if vm exists, wait for 15 mins before trying to delete
export VM_EXISTS=$(az vm list -o tsv --query "[?name=='$VM_NAME'].name")
if [ "$VM_EXISTS" != "" ];
then
echo "VM already exists from previous run. Waiting for 15 mins before deleting."
sleep 15m
az vm delete -n $VM_NAME -y
fi
# Create the VM. In the docs we will provide instructions to create a VM using az vm create -n $VM_NAME
az deployment group create --name $VM_NAME-$ENDPOINT_NAME --template-file endpoints/online/managed/vnet/setup_vm/vm-main.bicep --parameters vmName=$VM_NAME identityName=$IDENTITY_NAME vnetName=$VNET_NAME subnetName=$SUBNET_NAME
# command in script: az deployment group create --template-file endpoints/online/managed/vnet/setup/vm_main.bicep #identity name is hardcoded uai-identity
az vm run-command invoke -n $VM_NAME --command-id RunShellScript --scripts @endpoints/online/managed/vnet/setup_vm/scripts/vmsetup.sh --parameters "SUBSCRIPTION:$SUBSCRIPTION" "RESOURCE_GROUP:$RESOURCE_GROUP" "LOCATION:$LOCATION" "IDENTITY_NAME:$IDENTITY_NAME" "GIT_BRANCH:$GIT_BRANCH"
# build image
az vm run-command invoke -n $VM_NAME --command-id RunShellScript --scripts @endpoints/online/managed/vnet/setup_vm/scripts/build_image.sh --parameters "SUBSCRIPTION:$SUBSCRIPTION" "RESOURCE_GROUP:$RESOURCE_GROUP" "LOCATION:$LOCATION" "IDENTITY_NAME:$IDENTITY_NAME" "ACR_NAME=$ACR_NAME" "IMAGE_NAME:$IMAGE_NAME" "ENV_DIR_PATH:$ENV_DIR_PATH"
# create endpoint/deployment inside managed vnet
az vm run-command invoke -n $VM_NAME --command-id RunShellScript --scripts @endpoints/online/managed/vnet/setup_vm/scripts/create_moe.sh --parameters "SUBSCRIPTION:$SUBSCRIPTION" "RESOURCE_GROUP:$RESOURCE_GROUP" "LOCATION:$LOCATION" "IDENTITY_NAME:$IDENTITY_NAME" "WORKSPACE:$WORKSPACE" "ENDPOINT_NAME:$ENDPOINT_NAME" "ACR_NAME=$ACR_NAME" "IMAGE_NAME:$IMAGE_NAME" "ENDPOINT_FILE_PATH:$ENDPOINT_FILE_PATH" "DEPLOYMENT_FILE_PATH:$DEPLOYMENT_FILE_PATH" "SAMPLE_REQUEST_PATH:$SAMPLE_REQUEST_PATH"
# test the endpoint by scoring it
export CMD_OUTPUT=$(az vm run-command invoke -n $VM_NAME --command-id RunShellScript --scripts @endpoints/online/managed/vnet/setup_vm/scripts/score_endpoint.sh --parameters "SUBSCRIPTION:$SUBSCRIPTION" "RESOURCE_GROUP:$RESOURCE_GROUP" "LOCATION:$LOCATION" "IDENTITY_NAME:$IDENTITY_NAME" "WORKSPACE:$WORKSPACE" "ENDPOINT_NAME:$ENDPOINT_NAME" "SAMPLE_REQUEST_PATH:$SAMPLE_REQUEST_PATH")
# the scoring output for sample request should be [6141.267272547523, 6407.1333176127255]. We are validating if part of the number is available in the output (not comparing all the decimals to accomodate rounding discrepencies)
if [[ $CMD_OUTPUT =~ "6141" ]]; then
echo "Scoring works!"
else
echo "Error in scoring"
# delete the VM before exiting with error
az vm delete -n $VM_NAME -y --no-wait
# exit with error
exit 1
fi
### Cleanup
# <delete_endpoint>
az ml online-endpoint delete --name $ENDPOINT_NAME --yes --no-wait
# </delete_endpoint>
# <delete_vm>
az vm delete -n $VM_NAME -y --no-wait
# </delete_vm>