[Docs][Question] Example on how to adapt Multitask GP Regression to multi dimentional inputs.

How can one adapt the example on:

• GPyTorch Multitask_GP_Regression

which I can reproduce here for convenience:

train_x = torch.linspace(0, 1, 100)

train_y = torch.stack([
    torch.sin(train_x * (2 * math.pi)) + torch.randn(train_x.size()) * 0.2,
    torch.cos(train_x * (2 * math.pi)) + torch.randn(train_x.size()) * 0.2,
], -1)

to the case of 2 dimensional inputs, for example?

More specifically, I would like for the sine and cosine to have 2 inputs instead of 1.

for example, like this:

x = np.arange(-5, 5, 0.1)

y = np.arange(-5, 5, 0.1)

xx, yy = np.meshgrid(x, y)

z1 = np.sin(xx**2 + yy**2) / (xx**2 + yy**2)
z2 = np.cos(xx**2 + yy**2) / (xx**2 + yy**2)

[jacobrgardner]

This isn’t a dimensionality issue. Numpy’s default dtype is double while torch’s is float.

You need to do train_x = train_x.float() and train_y = train_y.float() to convert the data to fp32

[david-vicente]

I did not know about that. Thanks again!