Function test_AddLayer

numpy_ml/tests/test_nn.py:955–1008

(N=15)

def test_AddLayer(N=15):
    from numpy_ml.neural_nets.layers import Add
    from numpy_ml.neural_nets.activations import Tanh, ReLU, Sigmoid, Affine

    N = np.inf if N is None else N

    np.random.seed(12345)

    acts = [
        (Tanh(), nn.Tanh(), "Tanh"),
        (Sigmoid(), nn.Sigmoid(), "Sigmoid"),
        (ReLU(), nn.ReLU(), "ReLU"),
        (Affine(), TorchLinearActivation(), "Affine"),
    ]

    i = 1
    while i < N + 1:
        Xs = []
        n_ex = np.random.randint(1, 100)
        n_in = np.random.randint(1, 100)
        n_entries = np.random.randint(2, 5)
        for _ in range(n_entries):
            Xs.append(random_tensor((n_ex, n_in), standardize=True))

        act_fn, torch_fn, act_fn_name = acts[np.random.randint(0, len(acts))]

        # initialize Add layer
        L1 = Add(act_fn)

        # forward prop
        y_pred = L1.forward(Xs)

        # backprop
        dLdy = np.ones_like(y_pred)
        dLdXs = L1.backward(dLdy)

        # get gold standard gradients
        gold_mod = TorchAddLayer(torch_fn)
        golds = gold_mod.extract_grads(Xs)

        params = [(Xs, "Xs"), (y_pred, "Y")]
        params.extend(
            [(dldxi, "dLdX{}".format(i + 1)) for i, dldxi in enumerate(dLdXs)]
        )

        print("\nTrial {}".format(i))
        print("n_ex={}, n_in={}".format(n_ex, n_in))
        print("n_entries={}, act_fn={}".format(n_entries, str(act_fn)))
        for ix, (mine, label) in enumerate(params):
            assert_almost_equal(
                mine, golds[label], err_msg=err_fmt(params, golds, ix), decimal=1
            )
            print("\tPASSED {}".format(label))
        i += 1


def test_BatchNorm2D(N=15):