Merge pull request #667 from YashPandit4u/homogeneous-model-group-deployment

Adding sample homogeneous model group code.

Author: SudhiMurthy

model-group-deployment/homogeneous_model_group_deployment/README.md

@@ -0,0 +1,45 @@
+# Homogeneous Model Group Deployment
+
+Deploy multiple models together using OCI Data Science Model Groups for efficient multi-model serving from a single endpoint.
+
+## What This Notebook Covers
+
+- Building scikit-learn pipelines and registering models to the Model Catalog
+- Creating homogeneous model groups with multiple models
+- Deploying model groups to a single deployment endpoint
+- Performing inference using model keys
+- Live updates without downtime
+
+## Prerequisites
+
+- OCI Data Science project with Model Catalog access
+- Python packages: `ads`, `oci` (≥2.163.0), `scikit-learn`, `pandas`, `requests`
+- Environment variables: `PROJECT_OCID`, `NB_SESSION_COMPARTMENT_OCID`
+- OCI config file or resource principal authentication
+- Log group ID and log ID for deployment logging
+
+## Quick Start
+
+1. Open `homogeneous_model_group_deployment_example.ipynb` in your OCI Data Science notebook session
+2. Set `project_id`, `compartment_id`, `access_log_group_id`, and `log_id`
+3. Run cells sequentially
+
+## Inference Example
+
+Specify the model key in request headers to route to a specific model:
+
+```python
+headers = {
+    'Content-Type': 'application/json',
+    'model-key': 'svm1'  # Routes to specific model
+}
+response = requests.post(endpoint, json=data, auth=auth, headers=headers).json()
+```
+
+## Key Points
+
+- **Homogeneous groups**: Models share the same framework, scoring script, and runtime
+- **Model keys**: Each model has a unique inference key (e.g., "svm1", "svm2")
+- **Common artifact**: Requires a shared `score.py` and `runtime.yaml` for dynamic model loading
+- **Live updates**: Update deployments without downtime
+- **OCI SDK**: Some Model Group features require OCI SDK directly (not fully supported in ADS SDK yet)