Azure
A quick start guide for Featureform on Azure AKS.
This quickstart will walk through creating a few simple features, labels, and a training set using Postgres and Redis. We will use a transaction fraud training set.
- Python 3.7-3.10
- Kubectl
- Azure CLI
- An available domain/subdomain name
Install the Featureform SDK via Pip.
pip install featureform
Featureform uses gRPC which, in combination with the nginx ingress requires a fully qualified domain name.
export FEATUREFORM_HOST=<your_domain_name>
This step will provision a single node Kubernetes cluster with AKS
Login to the Azure CLI
az login
Create a resource group for the kubernetes cluster
az group create --name FeatureformResourceGroup --location eastus
Create a single node cluster for Featureform
az aks create --resource-group FeatureformResourceGroup --name FeatureformAKSCluster --node-count 1 --generate-ssh-keys
Add the cluster information to the kubeconfig as a the current context
az aks get-credentials --resource-group FeatureformResourceGroup --name FeatureformAKSCluster
kubectl get nodes
You should get a result like:
NAME STATUS ROLES AGE VERSION
aks-nodepool1-25554489-vmss000000 Ready agent 7m56s v1.24.6
We'll be installing three Helm Charts: Featureform, the Quickstart Demo, and Certificate Manager.
First we need to add the Helm repositories.
helm repo add featureform https://storage.googleapis.com/featureform-helm/
helm repo add jetstack https://charts.jetstack.io
helm repo update
Now we can install the Helm charts.
helm install certmgr jetstack/cert-manager \
--set installCRDs=true \
--version v1.8.0 \
--namespace cert-manager \
--create-namespace
helm install featureform featureform/featureform \
--set global.publicCert=true \
--set global.localCert=false \
--set global.hostname=$FEATUREFORM_HOST
helm install quickstart featureform/quickstart
Get the IP address of the ingress. It may take a minute or so to show.
kubectl get ingress
In your DNS provider create two records:
Key | Value | Record Type |
|---|---|---|
<your_domain_name> | <ingress_ip_address> | A |
<your_domain_name> | 0 issuewild "letsencrypt.org" | CAA |
This will allow the client to securely connect to the cluster by allowing the cluster to provision its own public IP address.
You can check when the cluster is ready by running
kubectl get cert
and checking that the status of the certificates is ready.
The Quickstart helm chart creates a Postgres instance with preloaded data, as well as an empty Redis standalone instance. Now that they are deployed, we can write a config file in Python.
definitions.py
import featureform as ff
redis = ff.register_redis(
name = "redis-quickstart",
host="quickstart-redis", # The internal dns name for redis
port=6379,
description = "A Redis deployment we created for the Featureform quickstart"
)
postgres = ff.register_postgres(
name = "postgres-quickstart",
host="quickstart-postgres", # The internal dns name for postgres
port="5432",
user="postgres",
password="password",
database="postgres",
description = "A Postgres deployment we created for the Featureform quickstart"
)
Once we create our config file, we can apply it to our Featureform deployment.
featureform apply definitions.py
We will create a user profile for us, and set it as the default owner for all the following resource definitions.
definitions.py
ff.register_user("featureformer").make_default_owner()
Now we'll register our user fraud dataset in Featureform.
definitions.py
transactions = postgres.register_table(
name = "transactions",
variant = "kaggle",
description = "Fraud Dataset From Kaggle",
table = "Transactions", # This is the table's name in Postgres
)
Next, we'll define a SQL transformation on our dataset.
definitions.py
@postgres.sql_transformation(variant="quickstart")
def average_user_transaction():
"""the average transaction amount for a user """
return "SELECT CustomerID as user_id, avg(TransactionAmount) " \
"as avg_transaction_amt from {{transactions.kaggle}} GROUP BY user_id"
Next, we'll register a passenger entity to associate with a feature and label.
definitions.py
user = ff.register_entity("user")
# Register a column from our transformation as a feature
average_user_transaction.register_resources(
entity=user,
entity_column="user_id",
inference_store=redis,
features=[
{"name": "avg_transactions", "variant": "quickstart", "column": "avg_transaction_amt", "type": "float32"},
],
)
# Register label from our base Transactions table
transactions.register_resources(
entity=user,
entity_column="customerid",
labels=[
{"name": "fraudulent", "variant": "quickstart", "column": "isfraud", "type": "bool"},
],
)
Finally, we'll join together the feature and label into a training set.
definitions.py
ff.register_training_set(
"fraud_training", "quickstart",
label=("fraudulent", "quickstart"),
features=[("avg_transactions", "quickstart")],
)
Now that our definitions are complete, we can apply it to our Featureform instance.
featureform apply definitions.py
Once we have our training set and features registered, we can train our model.
import featureform as ff
client = ff.ServingClient()
dataset = client.training_set("fraud_training", "quickstart")
training_dataset = dataset.repeat(10).shuffle(1000).batch(8)
for row in training_dataset:
print(row.features(), row.label())
We can serve features in production once we deploy our trained model as well.
import featureform as ff
client = ff.ServingClient()
fpf = client.features([("avg_transactions", "quickstart")], {"user": "C1410926"})
print(fpf)
Last modified 18d ago