test_positive.py

from basicgrad import Value
from basicgrad.nn import Neuron, MSELoss


def test_c2f():
    c2f = lambda x: float(1.8 * x + 32)
    model = Neuron(1, nonlin=False)
    model.parameters()
    X = [[Value(float(i))] for i in range(-5, 5)]
    y = [Value(c2f(i)) for i in range(-5, 5)]
    lr = 0.01
    while True:
        losss = []
        for x, yout in zip(X, y):
            pred = model(x)
            loss = (yout - pred) ** 2
            losss.append(loss)
        loss = sum(losss)
        if loss.data < 0.0001:
            break
        for pr in model.parameters():
            pr.grad = 0.0
        loss.backward()
        for pr in model.parameters():
            pr.data += -lr * pr.grad
    # print(model.parameters())
    assert sum([abs(model([Value(i)]).data - c2f(i)) for i in range(5, 10)]) / 5 < 0.01


def test_c2f_mse():
    c2f = lambda x: float(1.8 * x + 32)
    model = Neuron(1, nonlin=False)
    model.parameters()
    X = [[Value(float(i))] for i in range(-5, 5)]
    y = [Value(c2f(i)) for i in range(-5, 5)]
    lr = 0.01
    loss_fn = MSELoss()
    while True:
        preds = []
        for x in X:
            pred = model(x)
            preds.append(pred)
        loss = loss_fn(y, preds)
        if loss.data < 0.0001:
            break
        for pr in model.parameters():
            pr.grad = 0.0
        loss.backward()
        for pr in model.parameters():
            pr.data += -lr * pr.grad
    assert sum([abs(model([Value(i)]).data - c2f(i)) for i in range(5, 10)]) / 5 < 0.1