Help with variational Nearest neighborGP

[Chengwei94]

I was testing this tutorial https://docs.gpytorch.ai/en/latest/examples/07_Pyro_Integration/Pyro_GPyTorch_Low_Level.html but with variational nearest neighbor GP. The only thing I changed in the tutorial is in the init, but the posterior is just a straight line. Is there something I am doing wrong? Theres a lot of negative variance warning. I tried changing the kernel but results are around the same. Thanks!

``
from gpytorch.variational.nearest_neighbor_variational_strategy import NNVariationalStrategy

class PVGPRegressionModel(gpytorch.models.ApproximateGP):
def init(self, num_inducing=100, name_prefix="mixture_gp"):
self.name_prefix = name_prefix

    # Define all the variational stuff
    inducing_points = torch.linspace(0, 1, num_inducing)
    # variational_strategy = gpytorch.variational.VariationalStrategy(
    #     self, inducing_points,
    #     gpytorch.variational.CholeskyVariationalDistribution(num_inducing_points=num_inducing)
    # )
    inducing_points = inducing_points.reshape(-1, 1)
    m, d = inducing_points.shape
    self.m = m
    k = 64
    training_batch_size = 64
    self.k = k

    variational_distribution = gpytorch.variational.MeanFieldVariationalDistribution(m)


    variational_strategy = NNVariationalStrategy(self, inducing_points, variational_distribution, k=k,
                                                 training_batch_size=training_batch_size)

``