TorchDiagGaussian from logits

Ray Train
1

For box action spaces, are the logits in the form

[mean1, std1, mean2, std2, …]

or

[mean1, mean2, …, std1, std2, …]

?

2

Hi, can you give a bit more context on what you’re trying to build here and what Ray library you’re using? Thanks :slight_smile:

3

Hi. It seems like it has been some time, so you might have found it but here it is anyways:

it’s the second approach.

e.g.
output = torch.concat((means, stds), dim=-1)

4

@christina I am taking the torch model files produced from a ray train job, converting them to onnx files, and making a rust based inference program.

@iykim

Thanks!

5

Thank you @iykim ! Let me know if you have any other questions @H-Park :smiley: