Skip to content

README.md

Latest commit

 

History

History
25 lines (18 loc) · 1.8 KB

README.md

File metadata and controls

25 lines (18 loc) · 1.8 KB

🚀 MLOps Class: Image Classifier Deployment on GCP

This applied class project is the final MLOps deployment from the Data Science Retreat program. It demonstrates a robust, traceable pipeline for a ResNet Image Classifier (Fresh/Rotten Fruit classification).

⚠️ Deployment Status: The live application was successfully deployed on Google Cloud Platform (GCP) but has been deleted to manage costs. This repository maintains the complete, documented infrastructure for demonstrating pipeline and deployment expertise.

⚙️ Architecture and Key Technologies

Component Technology Purpose
Serving Endpoint FastAPI Provides a scalable, high-performance API for image prediction.
Data Validation Pydantic Ensures robust API request and response data integrity.
Artifact Management Weights & Biases (W&B) Dynamically fetches version-controlled model weights (best_model.pth) via the W&B API at container startup.
Model PyTorch / ResNet Optimized model loading and inference using torch.inference_mode().
Containerization Docker Ensures a portable, consistent runtime environment for deployment.
Cloud Target Google Cloud Platform (GCP) Repository showcases configuration for services like Cloud Run or Compute Engine.

📁 Repository Structure

The core application logic is contained within the app/ directory, following industry standards for containerized microservices.

  • app/main.py: Defines the FastAPI endpoint (/predict).
  • app/model.py: Crucial: Handles artifact fetching, environment configuration, and PyTorch model loading from W&B.
  • Dockerfile: Defines the build environment and startup command.
  • requirements.txt: Specifies all dependencies, including wandb and torch.