nbox-specimens

This repository contains follow along material and examples of using nbox. In case you do not have nbox installed:

pip install nbox

# call on CLI once to authenticate
nbx why

🍇 How to use

• [End to End] The stable-diffusion AI startup: [GPU] Train a Stable Diffusion model and serve it.

🌳 Projects

The projects here are used to demonstrate how to run different kinds of workloads and not the content of the operation.

• [Jobs] Sklearn Tests: Train a simple kmeans model and store the resulting chart in a Relic.
• [Jobs + LMAO] lmao: Use NBX-Monitoring projects to track, train and store your model giving a full end to end MLOps pipeline.
• [Serving] Pose Model: Deploy a mediapipe pose detection model on a kubernetes cluster (w/o YAML 😛)
• [Serving] FastAPI: Deploy a FastAPI server on a NBX-Serving
• [Jobs] DeepSpeed: [GPU] Train a MobileViT using DeepSpeed

Deprecated Examples

These examples will be migrated to the new format for deploying batch processes (serverless) and API endpoints.

• [Jobs & Deploy] A Job that deploys itself
• [Jobs] Credit Card Fraud Model: This will download a ~80MB csv file, preprocess and convert to a dataframe, do grid search for best model and securely transfer it back to a NBX-Build instance.
• [Jobs] GPU Jobs using faiss: Creates a GPU NBX-Job and runs the tests from facebookresearch/faiss
• [Jobs] gperc-insanity