feature: MLFlow E2E Example Notebook (5513)

[aviruthen]

Description

The notebook file v3-examples/ml-ops-examples/v3-mlflow-train-inference-e2e-example.ipynb already exists in the repository and contains a comprehensive MLflow E2E example. However, there are several issues that need to be fixed:

1. MLflow 3.x API incompatibility: The notebook uses registered_model.latest_versions which was deprecated/removed in MLflow 3.x. In MLflow 3.x, you should use client.search_model_versions() instead.
2. Endpoint invocation inconsistency: Step 7 uses raw boto3 sagemaker-runtime client to invoke the endpoint, but the existing e2e example (train-inference-e2e-example.ipynb) uses core_endpoint.invoke() which is the V3 pattern. The notebook should use the V3 pattern consistently.
3. Missing import json in Step 7: The json module is used in Step 7 but was only imported in Step 6 cell scope - it should work since cells share namespace, but better to keep consistent.
4. Session usage inconsistency: Step 1 uses Session.boto_region_name (class-level attribute) but other V3 examples instantiate a session first, then use session.boto_region_name (instance attribute). The existing local-training-example.ipynb instantiates a session first.
5. Missing sagemaker_session parameter: The ModelTrainer in Step 4 doesn't pass sagemaker_session, but the reference e2e notebook (train-inference-e2e-example.ipynb) does pass it.

Related Issue

Related issue: 5513

Changes Made

• v3-examples/ml-ops-examples/v3-mlflow-train-inference-e2e-example.ipynb

AI-Generated PR

This PR was automatically generated by the PySDK Issue Agent.

• Confidence score: 85%
• Classification: type: feature request
• SDK version target: V3

Merge Checklist

• Changes are backward compatible
• Commit message follows prefix: description format
• Unit tests added/updated
• Integration tests added (if applicable)
• Documentation updated (if applicable)

[sagemaker-bot]

🤖 AI Code Review

This PR fixes several issues in an MLflow E2E example notebook: updating to MLflow 3.x compatible APIs, using V3 SDK patterns (core_endpoint.invoke instead of raw boto3), adding sagemaker_session parameter, and fixing session instantiation. The changes are well-motivated and align with V3 SDK conventions, with a few minor suggestions.

[sagemaker-bot]

The import boto3 was moved here from Step 7, which is good for consolidation. However, per V3 architecture tenets, subpackages should avoid calling boto3 directly — the preferred pattern is to let Session() handle the boto session internally. Consider simplifying to:

from sagemaker.core.helper.session_helper import Session

sagemaker_session = Session()
AWS_REGION = sagemaker_session.boto_region_name

This avoids the explicit boto3 import entirely and is more consistent with V3's "Core as the foundation" tenet. The boto3 import was previously used in Step 7 for raw sagemaker-runtime client calls, which this PR correctly removes — so there's no remaining need for it.