huggingface-azure-acr

Overview

This project demonstrates an end-to-end MLOps pipeline using GitHub Actions for CI/CD to build and deploy a FastAPI application that leverages Hugging Face’s GPT-2 model for sentence completion. The deployment target is Azure Container Apps, with Docker images managed via Azure Container Registry (ACR).

• Tech Stack: Python, FastAPI, Hugging Face Transformers (GPT-2), Docker, Azure Container Registry, Azure Container Apps, GitHub Actions (CI/CD)
• Use case: Sentence completion using GPT-2 via a production-ready, scalable REST API

---

Features

• Automated CI/CD pipeline with GitHub Actions
• Dockerized FastAPI service for serving Hugging Face GPT-2
• Seamless deployment to Azure Container Apps from ACR
• Secure Azure authentication using repository secrets
• MLOps best practices for model deployment

---

Prerequisites

• Azure subscription with permissions to create resources
• Docker (for local builds/tests)
• Python 3.8+
• Azure CLI

---

Setup & Deployment

1. Clone the repository

git clone https://github.com/Muhammadyousafrana/huggingface-azure-acr.git
cd huggingface-azure-acr

2. Azure Credentials

To deploy via GitHub Actions, create AZURE_CREDENTIALS:

1. Open Azure Cloud Shell and run:

   az login

2. Create a service principal:

   az ad sp create-for-rbac \
     --name gha-deploy \
     --role contributor \
     --scopes /subscriptions/<subscription ID> \
     --sdk-auth

   • Replace <subscription ID> with your Azure Subscription ID.
   • Copy the JSON output.

3. In your GitHub repo, go to Settings > Secrets and variables > Actions, add a new secret named AZURE_CREDENTIALS, and paste the JSON output.

---

3. Build & Deploy

• On each push to main, GitHub Actions will:
  • Build the Docker image with FastAPI and GPT-2
  • Push to Azure Container Registry
  • Deploy to Azure Container Apps

You can monitor the deployment status via the GitHub Actions tab.

---

4. Logs

After deployment, view logs using Azure CLI:

az containerapp logs show \
  --name <container-app-name> \
  --resource-group <resource-group-name>

---

Usage

The deployed FastAPI service exposes a REST API for sentence completion using GPT-2.

Example request:

curl -X POST "https://<your-app-url>/complete" -H "Content-Type: application/json" -d '{"prompt": "Once upon a time"}'

Response:

{
  "completion": "Once upon a time, there was a..."
}

---

Contributing

Contributions are welcome! Please open issues or submit pull requests for improvements.

---

License

MIT License

---

Contact

For questions, open an issue or contact the repo owner.

---