tannydobreva / sagemaker-workshop-labs

Machine Learning Workshop

Prerequisites

Sign in to the AWS console

• Step 1. Paste the hash login URL associated to your name in a browser

• Step 2. Click “AWS Console”, then “Open AWS Console”

Open Amazon SageMaker Studio

• Step 3. Navigate to Amazon SageMaker service

• Step 4. Create Amazon SageMaker Studio session in Oregon (us-west-2) region.

  Select "Quick start" with "Create a new role" in the "Execution Role" section, then click "Submit". This will take about 2-3 mins.

  Once its completed and the Amazon SageMaker Studio domain has status "Ready", select "Open Studio".

Clone GitHub repository within the Amazon SageMaker Studio

• Step 5. Within Amazon SageMaker Studio, go to the File → New → Terminal and then run following command

git clone https://github.com/veerathp/sagemaker-workshop-labs.git

Hands-on labs

Use Python 3 (Data Science) kernel to run all labs below into SageMaker Studio. For SageMaker notebook instances use conda_python3.

Lab 1: Build, Train and Deploy ML models

Build, Train and Deploy Machine Learning models using Amazon SageMaker / DeerAR

• Open and run the Notebook Instances in Amazon SageMaker Studio located in /01-DeepAR/deepar_chicago_traffic_violations.ipynb

Lab 2: Hyper Parameter Optimization

Hyper Parameter Optimization using Amazon SageMaker. The first lab will take about 20-30mins to run, once completed move to the second notebook to analyzing the training jobs.

• Open and run the Notebook Instances in Amazon SageMaker Studio located in /04-sagemaker-hpo-xgboost/hpo_xgboost_direct_marketing_sagemaker_python_sdk.ipynb

• Cont. Open and run the Notebook Instances in Amazon SageMaker Studio located in /04-sagemaker-hpo-xgboost/HPO_Analyze_TuningJob_Results.ipynb. Make sure you replace the name of the training job

Lab 3: Running experiments

Running experiments and manage multiple trials

• Open and run the Notebook Instance in Amazon SageMaker Studio located in /02-customer-churn/xgboost_customer_churn_studio.ipynb

Lab 4: Unsupervised ML algorithms

Build, Train and Deploy Machine Learning models for anomaly detection

• Open and run the Notebook Instance in Amazon SageMaker Studio located in /03-unsupervised-random-cut-forest/random_cut_forest.ipynb

Lab 5: Multi Model endpoints

Deploying Machine Learning models using Multi Model endpoints

• Open and run the Notebook Instance in Amazon SageMaker Studio located in /06-sagemaker-mme/xgboost_multi_model_endpoint_home_value.ipynb

Additional Resources

• Amazon SageMaker
• Amazon SageMaker Examples
• Amazon SageMaker Studio
• Amazon SageMaker Autopilot
• Host Multiple Models with Multi-Model Endpoints