FastAPI for model serving in ML Python - Model Pipeline Trace

Model Pipeline - FastAPI for model serving

This pipeline shows how a trained machine learning model is served using FastAPI. It takes input data from a user, processes it, runs the model to get predictions, and returns the results quickly through a web API.

Data Flow - 4 Stages

1Input Data Received

1 row x 4 features→Receive JSON data via FastAPI POST request→1 row x 4 features

{"sepal_length": 5.1, "sepal_width": 3.5, "petal_length": 1.4, "petal_width": 0.2}

↓

2Data Preprocessing

1 row x 4 features→Convert JSON to numpy array and scale features→1 row x 4 features

[[0.22, 0.42, 0.11, 0.04]]

↓

3Model Prediction

1 row x 4 features→Run input through trained classifier→1 row x 3 classes

[[0.95, 0.03, 0.02]]

↓

4Output Formatting

1 row x 3 classes→Convert prediction probabilities to class label→1 prediction label

"setosa"

Training Trace - Epoch by Epoch

Loss
0.65 |*****
0.45 |****
0.30 |***
0.20 |**
0.15 |*

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.65	0.70	Model starts learning basic patterns
2	0.45	0.82	Loss decreases, accuracy improves
3	0.30	0.90	Model converging well
4	0.20	0.95	High accuracy, low loss
5	0.15	0.97	Training stabilizes with good performance

Prediction Trace - 4 Layers

Layer 1: Receive JSON input

Layer 2: Preprocessing

Layer 3: Model Prediction

Layer 4: Output Formatting

Model Quiz - 3 Questions

What shape does the input data have when received by FastAPI?

A 100 rows x 4 features

B 1 row x 3 classes

C 1 row x 4 features

D 4 rows x 1 feature

Key Insight

Serving a machine learning model with FastAPI involves receiving input data, preprocessing it to match the model's needs, running the model to get predictions, and returning the results quickly. This pipeline ensures smooth interaction between users and the model in real time.