jaeyow

# Pre-process and model building
name: F1 Prediction MLOps

on:
  workflow_dispatch

jobs:
  build:

    runs-on: ubuntu-latest

jaeyow

from metaflow import FlowSpec, step, current
from comet_ml import API, Experiment
import os
import random

try:
    from dotenv import load_dotenv
    load_dotenv(verbose=True, dotenv_path='.env')
except:
    print("No dotenv package")

jaeyow

@step
def model_training(self):
  """
  Model training
  - now training starts, first we specify the Docker image for the required algorithm, in this case linear learner
  - create an estimator with the specified parameters, 
  - set the static hyperparameters, and SageMaker will automatically calculate those set as 'auto'
  - calling fit() starts the training process, upto the specified number of epochs
  - the save the model name and location for the next steps
  - take note that we have to specify an instance for training, which may be different from the endpoint instance

jaeyow

def getLabelsOfImage(bucket_name, number_labels, image):
    # Use boto3 call detect_labels to get Amazon Rekognition labels
    client = boto3.client("rekognition")
    response = client.detect_labels(
        Image={"S3Object": {"Bucket": bucket_name, "Name": image}},
        MaxLabels=number_labels,
        MinConfidence=98
    )
    return response["Labels"]

jaeyow

def getCustomLabelsOfImage(model_arn, bucket_name, image):
    # Use boto3 call detect_custom_labels API call to get Rekognition Custom Labels
    client = boto3.client("rekognition")
    response = client.detect_custom_labels(
        ProjectVersionArn=model_arn,
        Image={"S3Object": {"Bucket": bucket_name, "Name": image}},
        MinConfidence=98,
    )
    return response["CustomLabels"]

jaeyow

import pandas as pd
import boto3
from generators.generate_users_json import generate_users_json

bucket = "cevo-shopping-demo"
users_filename = "./generators/users.csv"

generate_users_json()

boto3.Session(profile_name=<aws-profile-replace-me>, region_name="ap-southeast-2").resource(

jaeyow

import boto3
from generators.generate_items_and_interactions_personalize import generate_interactions

bucket = "cevo-shopping-demo"
interactions_filename = "./generators/interactions.csv"

generate_interactions()

boto3.Session(profile_name=<aws-profile-replace-me>, region_name="ap-southeast-2").resource(
    "s3"

jaeyow

import boto3
from generators.generate_items_and_interactions_personalize import generate_user_items

bucket = "cevo-shopping-demo"
products_filename = "./generators/items.csv"

generate_user_items()

boto3.Session(profile_name=<aws-profile-replace-me>, region_name="ap-southeast-2").resource(
    "s3"

jaeyow

def get_recommended_items(user_id):
    ssm = boto3.client("ssm")
    recommenderArn = ssm.get_parameter(
        Name="/cevo-shopping-demo/recommender/arn-retaildemostore-recommended-for-you",
        WithDecryption=False,
    )
    response = personalizeRuntime.get_recommendations(
        recommenderArn=recommenderArn["Parameter"]["Value"],
        userId=user_id,
        numResults=30,

jaeyow

def get_popular_items(user_id="x"):
    ssm = boto3.client("ssm")
    recommenderArn = ssm.get_parameter(
        Name="/cevo-shopping-demo/recommender/arn-retaildemostore-popular-items",
        WithDecryption=False,
    )
    response = personalizeRuntime.get_recommendations(
        recommenderArn=recommenderArn["Parameter"]["Value"],
        userId=user_id,
        numResults=100,