example1.py

import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
from tqdm import tqdm


BITS = 100
NUMBERS = 1000


def int_to_binary(n, bits=BITS):
    return np.array([int(bit) for bit in np.binary_repr(n, width=bits)])


num_array = np.arange(1, NUMBERS + 1)
X = np.array([int_to_binary(i) for i in num_array], dtype=np.float32)
Y = np.array([1 if i % 2 == 0 else 0 for i in num_array], dtype=np.float32).reshape(
    -1, 1
)
X_tensor = torch.tensor(X, dtype=torch.float32)
Y_tensor = torch.tensor(Y, dtype=torch.float32)


class BinaryClassifier(nn.Module):
    def __init__(self, input_size=BITS):
        super(BinaryClassifier, self).__init__()
        self.fc1 = nn.Linear(input_size, 8)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(8, 1)
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        x = self.fc1(x)
        x = self.relu(x)
        x = self.fc2(x)
        x = self.sigmoid(x)
        return x


criterion = nn.BCELoss()
model = BinaryClassifier()
optimizer = optim.Adam(model.parameters(), lr=0.001)


print("Training:")
with tqdm(range(NUMBERS), desc="", position=0, leave=True) as progress_bar:
    for epoch in range(NUMBERS):
        model.train()

        optimizer.zero_grad()

        outputs = model(X_tensor)
        loss = criterion(outputs, Y_tensor)
        loss.backward()

        optimizer.step()

        preds = (outputs.detach().cpu().numpy() > 0.5).astype(int).flatten()
        accuracy = (preds == Y.flatten()).mean() * 100

        progress_bar.update()
        progress_bar.set_description(
            f"loss: {loss.item():6.2f} accuracy: {accuracy:5.2f}%"
        )


def test_model(number):
    model.eval()
    with torch.no_grad():
        number_tensor = torch.tensor(
            np.array(int_to_binary(number)), dtype=torch.float32
        )
        prediction = model(number_tensor).item()
        pred = "even" if prediction > 0.5 else "odd"
        print(f"Number {number} -> {pred}")


test_numbers = [17, 44, 62, 503, 12321]
print("Prediction:")
for n in test_numbers:
    test_model(n)