How to use the pyriemann.spatialfilters.BilinearFilter function in pyriemann

def test_BilinearFilter():
    """Test Bilinear filter"""
    n_trials = 90
    X = generate_cov(n_trials, 3)
    labels = np.array([0, 1, 2]).repeat(n_trials // 3)
    filters = np.eye(3)
    # Test Init
    bf = BilinearFilter(filters)
    assert_false(bf.log)
    assert_raises(TypeError, BilinearFilter, 'foo')
    assert_raises(TypeError, BilinearFilter, np.eye(3), log='foo')

    # test fit
    bf = BilinearFilter(filters)
    bf.fit(X, labels % 2)

    # Test transform
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])
    assert_raises(TypeError, bf.transform, 'foo')
    assert_raises(ValueError, bf.transform, X[:, 1:, :])  # unequal # of chans
    bf.log = True
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0]])

    filters = filters[0:2, :]
    bf = BilinearFilter(filters)
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])

def test_BilinearFilter():
    """Test Bilinear filter"""
    n_trials = 90
    X = generate_cov(n_trials, 3)
    labels = np.array([0, 1, 2]).repeat(n_trials // 3)
    filters = np.eye(3)
    # Test Init
    bf = BilinearFilter(filters)
    assert_false(bf.log)
    assert_raises(TypeError, BilinearFilter, 'foo')
    assert_raises(TypeError, BilinearFilter, np.eye(3), log='foo')

    # test fit
    bf = BilinearFilter(filters)
    bf.fit(X, labels % 2)

    # Test transform
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])
    assert_raises(TypeError, bf.transform, 'foo')
    assert_raises(ValueError, bf.transform, X[:, 1:, :])  # unequal # of chans
    bf.log = True
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0]])

    filters = filters[0:2, :]

def test_BilinearFilter():
    """Test Bilinear filter"""
    n_trials = 90
    X = generate_cov(n_trials, 3)
    labels = np.array([0, 1, 2]).repeat(n_trials // 3)
    filters = np.eye(3)
    # Test Init
    bf = BilinearFilter(filters)
    assert_false(bf.log)
    assert_raises(TypeError, BilinearFilter, 'foo')
    assert_raises(TypeError, BilinearFilter, np.eye(3), log='foo')

    # test fit
    bf = BilinearFilter(filters)
    bf.fit(X, labels % 2)

    # Test transform
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])
    assert_raises(TypeError, bf.transform, 'foo')
    assert_raises(ValueError, bf.transform, X[:, 1:, :])  # unequal # of chans
    bf.log = True
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0]])

# test fit
    bf = BilinearFilter(filters)
    bf.fit(X, labels % 2)

    # Test transform
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])
    assert_raises(TypeError, bf.transform, 'foo')
    assert_raises(ValueError, bf.transform, X[:, 1:, :])  # unequal # of chans
    bf.log = True
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0]])

    filters = filters[0:2, :]
    bf = BilinearFilter(filters)
    Xt = bf.transform(X)
    assert_array_equal(Xt.shape, [len(X), filters.shape[0], filters.shape[0]])

raise ValueError("Data and filters dimension must be compatible.")

        X_filt = numpy.dot(numpy.dot(self.filters_, X), self.filters_.T)
        X_filt = X_filt.transpose((1, 0, 2))

        # if logvariance
        if self.log:
            out = numpy.zeros((len(X_filt), len(self.filters_)))
            for i, x in enumerate(X_filt):
                out[i] = numpy.log(numpy.diag(x))
            return out
        else:
            return X_filt


class CSP(BilinearFilter):
    """Implementation of the CSP spatial Filtering with Covariance as input.

    Implementation of the famous Common Spatial Pattern Algorithm, but with
    covariance matrices as input. In addition, the implementation allow
    different metric for the estimation of the class-related mean covariance
    matrices, as described in [3].

    This implementation support multiclass CSP by means of approximate joint
    diagonalization. In this case, the spatial filter selection is achieved
    according to [4].

    Parameters
    ----------
    nfilter : int (default 4)
        The number of components to decompose M/EEG signals.
    metric : str (default "euclid")