def test_no_change(self, mode, padding_mode):
# scale=1
n_samples, n_channels, lookback, n_assets = 2, 3, 4, 5
X = torch.rand(n_samples, n_channels, lookback, n_assets)
scale = torch.ones(n_samples, dtype=X.dtype)
layer = Zoom(mode=mode, padding_mode=padding_mode)
x_zoomed = layer(X, scale)
assert torch.allclose(x_zoomed, X)
def test_basic(self, Xy_dummy, mode, padding_mode):
X, _, _, _ = Xy_dummy
dtype, device = X.dtype, X.device
n_samples, _, _, n_assets = X.shape
layer = Zoom(mode=mode, padding_mode=padding_mode)
scale = torch.rand(n_samples, dtype=dtype, device=device)
x_zoomed = layer(X, scale)
assert torch.is_tensor(x_zoomed)
assert x_zoomed.shape == X.shape
assert x_zoomed.dtype == X.dtype
assert x_zoomed.device == X.device
def test_equality_with_warp(self):
n_samples, n_channels, lookback, n_assets = 2, 3, 4, 5
X = torch.rand(n_samples, n_channels, lookback, n_assets)
scale = torch.ones(n_samples, dtype=X.dtype) * 0.5
tform = torch.stack(n_samples * [torch.linspace(0, end=1, steps=lookback)], dim=0)
layer_zoom = Zoom()
layer_warp = Warp()
x_zoomed = layer_zoom(X, scale)
x_warped = layer_warp(X, tform)
assert torch.allclose(x_zoomed, x_warped)
lookback, n_assets = 40, 50
scale_grid = [1, 0.5, 2]
padding_grid = ['zeros', 'reflection']
dtype = torch.float
x = torch.as_tensor(np.stack(n_assets *
[-sin_single(lookback, freq=1 / lookback)],
axis=1),
dtype=dtype)
x = x[None, None, ...] # add sample and channel dimension
fig, axs = plt.subplots(len(padding_grid),
len(scale_grid),
sharex=True,
sharey=True)
for r, padding_mode in enumerate(padding_grid):
for c, scale in enumerate(scale_grid):
layer = Zoom(padding_mode=padding_mode)
x_out = layer(x, torch.ones(1) * scale)
axs[r, c].imshow(x_out.numpy()[0, 0])
axs[r, c].set_title('scale={}\npad={}'.format(scale, padding_mode))
plt.setp(axs[-1, :], xlabel='Assets')
plt.setp(axs[:, 0], ylabel='Time')
fig.subplots_adjust(hspace=1)
def test_n_parameters(self):
n_parameters = sum(p.numel() for p in Zoom().parameters()
if p.requires_grad)
assert n_parameters == 0