Simple MNIST digit classifier using TinyGrad with clean terminal UI. Supports training, model saving/loading, and prediction on custom images.
To get started, clone this repository:
git clone https://github.com/alouiadel/tinygrad-mnist.git
cd tinygrad-mnist
# Setup
python -m venv tiny-venv
source tiny-venv/bin/activate # or `tiny-venv\Scripts\activate` on Windows
pip install -r requirements.txt
# Train
python main.py
# Predict
# Place digit images (PNG/JPG) in the 'pics' folder
python main.py # Select option to load model and predict
- 🎯 Train on MNIST dataset with live progress display
- 💾 Save/load trained models
- 🔍 Predict digits from custom images
- 🎨 Clean terminal UI with colored output
Tinygrad documentation(initial source): Training an MNIST Classifier.