Sampling from posterior with 3-D tensor

[Jimmyc776]

Good evening.

I'm working on a model with simulated data stored in a 3-D tensor and a last dimension of 3. I have successfully setup SNPE inference and amortized trained with an embedding net. The next thing I wish to do is sample from this posterior using "posterior.sample((sampleSize,), x=tensor)", where tensor is the simulated data with last dimension of 3. However, I get a value error with a very long message. Attached is a screenshot of the whole error message. What, if anything, can I do to this tensor to fix my issue?

Thanks!

[michaeldeistler]

Hi there,

thanks for reporting this! I tried reproducing the issue but all seems fine to me. Could you test whether the following works for you:

import torch
from torch import ones, zeros, eye, float32, tensor, as_tensor, randn, rand, randn_like
from torch.nn import Module
from torch import nn

from sbi.inference import NPE
from sbi.utils import BoxUniform
from sbi.neural_nets import posterior_nn


prior = BoxUniform(-ones(3), ones(3))
theta = prior.sample((100,))
x = torch.stack([theta, theta, theta], axis=1)
x = torch.stack([x, x, x], axis=1)
x += randn_like(x) * 0.1
print(f"theta.shape {theta.shape}")  # torch.Size([100, 3])
print(f"x.shape {x.shape}")  # torch.Size([100, 3, 3, 3])

x_o = x[0]  # x[:1] also works.
print(f"x_o.shape {x_o.shape}")  # torch.Size([3, 3, 3])

class Embedding3D(Module):
    def __init__(self):
        super().__init__()
        self.net = nn.Linear(3**3, 20)

    def __call__(self, x):
        return self.net(torch.reshape(x, (-1, 3**3)))

embedding_net = Embedding3D()
density_estimator = posterior_nn("maf", embedding_net=embedding_net)

inference = NPE(prior, density_estimator=density_estimator)
_ = inference.append_simulations(theta, x).train()

posterior = inference.build_posterior()
samples = posterior.sample((100,), x=x_o)
print(f"Posterior samples.shape: {samples.shape}")  # torch.Size([100, 3])

If this works, could you share a minimal example to reproduce your issue?

Thanks!
Michael

[Jimmyc776]

Issue resolved! Thank you very much for your response.

I don't have a good code example to share. I realized, after working through the code you provided, that I had tried to use the wrong x_o tensor.

-- Jimmy

[michaeldeistler]

Glad to hear that!