cereal-ml

Simple PyTorch-based calorie prediction using cereal nutrition data.

Predict cereal calories from nutritional features using:

• CLI commands
• Python API calls
• Trained PyTorch models

---

Features

• Train calorie prediction model from cereal nutrition data
• Save/load PyTorch models
• CLI predictions
• Python API predictions
• Loss plotting
• MAE evaluation
• Simple custom feature parser

---

Installation

Clone the repo:

git clone https://github.com/poti1/cereal-ml.git
cd cereal-ml

Install dependencies:

pip install -r requirements.txt

---

Training

Train the model:

./cereal.py --train

Example output:

Example training metrics:

Epoch 1/10  loss: 104.8351  val_loss: 111.3147
...
Epoch 10/10 loss: 25.8951  val_loss: 27.3140

mae: 28.50

---

CLI Prediction

Run predictions directly from the command line:

./cereal.py --data 'sugars=3, protein=5'

Example:

---

Python API Usage

Use the predictor directly in Python:

import cereal

cereal.predict('fat=10, sodium=20, carbo=15')

Example:

---

Example Predictions

Input	Predicted Calories
sugars=3, protein=5	71
fat=10, sodium=20, carbo=15	268

---

Plotting Loss

Generate a training loss plot:

./cereal.py --plot

This creates:

loss.png

---

Model Output

The trained model is saved as:

cereal.pt

---

Tech Stack

• Python
• PyTorch
• pandas
• matplotlib
• scikit-learn

---

Goals

This project was built to practice:

• ML training pipelines
• Regression models
• Feature parsing
• Model serialization
• CLI + Python interfaces
• End-to-end ML workflows

---

License

MIT