fpaupier / sidecar-pattern_tf-serving

Introducing model_poller, a sidecar container for tensorflow/serving.

Machine Learning Engineer & Technical Leader

🛠 Core Competencies:

• Machine Learning Workflow Design & Optimization
  • Natural Language Processing (NLP) & Computer Vision (CV)
  • Model training, deployment & monitoring - PyTorch, LabelStudio, MLFlow, DVC
• Distributed Data Processing & Scalable Architectures (docker, docker-compose, kubernetes)
• Cloud Computing (Scaleway, AWS, GCP)
• Full-Stack Development
  • Backend (Python, FastAPI, gRPC)
  • Frontend (Vanilla JS/HTML, React)

📚 Education: Centrale Nantes - Master of Science: Signal Processing & Statistics.

🗂 Professional Journey:

• ML Engineer & Eng. manager @ French DoD
• Freelance ML Engineer
• ML Engineer @ Schneider Electric, Boston, MA
• Data Engineer @ Kpler, Paris

🔍 Learning: Currently looking at Robotics, Evolutionary Computing & Metaheuristics, Conflict-free Replicated Data Types (CRDTs)

Featured Content 📝

• 📕 Blog Posts:

  • Large Language Model (LLM) Finetuning & Evaluation - Practical Insights
  • Apache NiFi - Surf on Your Dataflow
  • Multiprocessing with Python & gRPC
  • Telegram Bot in Go

• 🎙 Podcast - Post Mortem Podcast: Exploring the intricacies of system outages, causes, and solutions. Available on Apple Podcasts, Spotify, and Deezer

Tech Stack 🛠️

My favorite tool is the one that's right for the job at hand. Still, I've developed a particular affinity for some.

Connect 🔗

---

---

Credits

Banner designed using the github profile header generator, octocat shaped with the myOctocat creator.

sidecar-pattern_tf-serving's Issues

Isn't model polling already a part of TF Serving?

I was reading through the documentation of the TF Serving architecture explicitly. It mentions that utilities for polling a file-system are made available out of the box.

New Sources could support new filesystems, cloud offerings and algorithm backends. TensorFlow Serving provides some common building blocks to make it easy & fast to create new sources. For example, TensorFlow Serving includes a utility to wrap polling behavior around a simple source. Sources are closely related to Loaders for specific algorithms and data hosting servables.

Did you try using these instead of Pub/Sub?