bedrock-examples

A repository to host examples for notebooks, pipelines, batch scoring and deployment on Bedrock

This repository is home to the following types of examples:

• End-to-end
• Component

hcl template

A copy of the hcl file is provided as a template.

End-to-end

Regression

This example covers the following concepts:

1. Trains a regression model on bedrock
2. Log metrics for regression using the bedrock client
3. Customise logging of feature distributions
4. Log model explanabilty and fairness for regression on bedrock
5. Serve a trained regression model using bedrock express

Binary class

This example covers the following concepts:

1. Set up a Bedrock training pipeline, either on Google Cloud or AWS
2. Monitor the training
3. Deploy a model endpoint in HTTPS
4. Query the endpoint
5. Monitor the endpoint API metrics

Credit risk

This example covers the following concepts:

1. Set up a Bedrock training pipeline
2. Log training-time feature and inference distributions
3. Log model explainability and fairness metrics
4. Check model explainability and fairness from Bedrock web UI
5. Deploy a model endpoint in HTTPS with logging feature and inference distributions
6. Monitor the endpoint by simulating a query stream

Multiclass

This example covers the following concepts:

1. Train a multiclass classification machine learning model
2. Log metrics for multiclass classfication
3. Log ROC, PR, and confusion matrices by micro-averaging on all classes
4. Customise logging of feature distributions
5. Log predictions for multiclass classification
6. Log model explainabilty and fairness for multiclass
7. Serving code for multiclass classification using bedrock express

CV demo

This example covers the following concepts:

1. Train a model with GPU on Bedrock
2. Deploy a model endpoint
3. Deploy a Streamlit app

Component

Miniconda

This example covers the following concepts:

1. Install conda packaages
2. Activate a conda environment on bedrock

PySpark step

This example covers the following concepts:

1. Write hcl with Spark configuration

Use GPU

This example covers the following concepts:

1. Deploy a GPU instance

Batch scoring pipeline

This example covecrs the following concepts:

1. Set up a Bedrock batch scoring pipeline on AWS

Serve with gRPC

This example covers the following concepts:

1. Deploy a gRPC endpoint on Google Cloud

NLP demo

This example covers the following concepts:

1. Fine-tune a pre-trained model
2. Use Captum for explainability

Documentation

Refer to the documentation for more details.