Loading a pre-trained model

Deploying AI into Production with FastAPI

Matt Eckerle

Software and Data Engineering Leader

Current structure

Flowchart showing ML prediction with FastAPI

Challenge with loading models

A diagram explaining request/response lifecycle

A frustrated man waiting for API response

Load models before the request

Flowchart showing ML prediction with FastAPI

Loading the model

from fastapi import FastAPI

sentiment_model = None

def load_model():
    global sentiment_model
    sentiment_model = SentimentAnalyzer("trained_model.joblib")
    print("Model loaded successfully")

load_model()

Model loaded successfully

FastAPI lifespan event

A flow diagram explaining FastAPI lifespan

FastAPI lifespan event

from contextlib import asynccontextmanager

@asynccontextmanager
async def lifespan(app: FastAPI):
    # Startup: Load the ML model
    load_model()
    yield

app = FastAPI(lifespan=lifespan)

Health checks

@app.get("/health")
def health_check():
    if sentiment_model is not None:
        return {"status": "healthy", 
                "model_loaded": True}
    return {"status": "unhealthy", 
            "model_loaded": False}

Curl command:

curl -X GET \
  "http://localhost:8080/health" \
  -H "accept: application/json"

Output:

{
  "status": "healthy",
  "model_loaded": true
}

Let's practice!

Deploying AI into Production with FastAPI