This starter is tuned for a Mac M1/M2 on CPU (and can optionally use Metal GPU). It sets up a baseline CIFAR-10 model and exports TFLite models for edge experiments.

# 0) Create & activate a clean env (choose one)
# With conda/mamba:
conda create -n mso python=3.10 -y && conda activate mso
# OR with venv:
python3 -m venv .venv && source .venv/bin/activate

# 1) Install deps
pip install -U pip
pip install -r requirements.txt

# 2) Run baseline (CPU by default)
python part1_baseline.py

# Artifacts will appear under:
# - models/baseline_model.keras
# - exports/*.tflite
# - reports/baseline_metrics.json
# - charts/history_accuracy_loss.png

If you don't want to use the Apple GPU, do nothing (CPU is default). If you do want GPU acceleration, keep tensorflow-metal installed—no extra code changes needed.

• part1_baseline.py: trains CIFAR-10 baseline with callbacks, evaluates, exports TFLite, logs metrics.
• requirements.txt: minimal dependencies for Apple Silicon.
• reports/, models/, exports/, charts/: output folders.