Digit count array:

[1309, 1576, 948, 1273, 1186, 1301, 1222, 745, 1006, 1434]

Sum: 12000, files: 12000

Чтобы установить зависимости на другом компьютере:

pip install -r requirements.txt

1. Key Features

• Convolutional Neural Network trained on MNIST
• All randomness eliminated: full determinism
• Saved model (mnist_model.h5) ensures identical output on every run
• CPU-only execution
• No non-deterministic TensorFlow kernels
• Output is 100% reproducible across machines and operating systems

---

2. Deterministic Execution Guarantee

This solution guarantees identical results on every run due to:

1. PYTHONHASHSEED, numpy, random seeds fixed to 42
2. TF_DETERMINISTIC_OPS=1
3. TF_ENABLE_ONEDNN_OPTS=0
4. XLA disabled
5. Model batch order is not shuffled (shuffle=False)
6. Model is trained once and saved
7. All subsequent runs reuse the saved model