train.py

from engine import Value
from neuron import Neuron
from layer import Layer
from loss_functions import MSELoss
from optimizer import SGD
from MLP import MLP
from LRScheduler import StepLRScheduler
import random

mlp = MLP(4, [128, 64, 1])
loss_fn = MSELoss()
scheduler = StepLRScheduler(lr=0.1)
optimizer = SGD(mlp.parameters(), scheduler = scheduler)



train_X = [[Value(random.uniform(-1,1)) for _ in range(4)] for _ in range(10)]
train_Y = [[Value(random.uniform(-1,1))] for _ in range(10)]

epochs = 30

for epoch in range(epochs):
    total_loss = 0

    for x, y_true in zip(train_X, train_Y):
        y_pred = mlp(x)


        loss = loss_fn(y_pred, y_true)
        total_loss += loss.data

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

    avg_loss = total_loss / len(train_X)
    print(f"Epoch {epoch + 1}/{epochs}, Loss: {avg_loss:.4f}")

print(mlp)