Why architecture choices affect scalability in Prompt Engineering / GenAI - Model Pipeline Impact

Model Pipeline - Why architecture choices affect scalability

This pipeline shows how choosing different model architectures impacts how well a machine learning model can handle more data and bigger tasks without slowing down or using too much memory.

Data Flow - 6 Stages

1Input Data

10000 rows x 20 features→Raw data collected for training→10000 rows x 20 features

Each row has 20 numbers representing different measurements

↓

2Preprocessing

10000 rows x 20 features→Normalize features to range 0-1→10000 rows x 20 features

Feature values scaled so all are between 0 and 1

↓

3Feature Engineering

10000 rows x 20 features→Add polynomial features (degree 2)→10000 rows x 230 features

New features created by multiplying pairs of original features

↓

4Model Architecture Choice

10000 rows x 230 features→Select between small or large neural network→10000 rows x 230 features

Small model: 2 layers with 50 neurons each; Large model: 5 layers with 200 neurons each

↓

5Training

10000 rows x 230 features→Train model with chosen architecture→Trained model parameters

Weights adjusted to reduce error on training data

↓

6Prediction

New data 1 row x 230 features→Model predicts output→1 row x 1 output

Model outputs a number representing predicted value

Training Trace - Epoch by Epoch

Epochs
1 |***************
5 |***************
10|********************
15|***********************
20|*************************
Loss
0.85 0.60 0.40 0.30 0.25

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.85	0.55	Starting training with high loss and low accuracy
5	0.60	0.70	Loss decreasing, accuracy improving steadily
10	0.40	0.82	Model learning well, loss much lower
15	0.30	0.88	Good convergence, accuracy nearing 90%
20	0.25	0.91	Training stabilizes with low loss and high accuracy

Prediction Trace - 3 Layers

Layer 1: Input Layer

Layer 2: Hidden Layers (Large Model)

Layer 3: Output Layer

Model Quiz - 3 Questions

Why does a larger model architecture affect scalability?

A It reduces the number of features

B It always makes the model less accurate

C It uses more memory and takes longer to train

D It makes data preprocessing faster

Key Insight

Choosing a larger or more complex model architecture can improve accuracy but requires more memory and time, affecting how well the model scales to bigger datasets or tasks.