Matthew Lyon

@Tallal64 this does not resolve the issue for me. Instead, i set `SVGstyle={{ display: 'inline-block' }}` which resolved the issue.

Of course, my apologies for not including that info. I'm attempting to use the [Laplace framework](https://github.com/AlexImmer/Laplace/blob/main/laplace/curvature/backpack.py) for a model, in my current implementation it's using the `DiagGGNExact` extension.

Thank you for pointing me in the right direction. I'll take a look at the second order extension example and see if I can't implement something myself. Unfortunately the Laplace...

> This means that your model must entirely consist of layers known to BackPACK and it must be provided as a collection of modules without overwriting forward It doesn't seem...

They are somewhat complex. I think being able to break the problem down into smaller parts would be a much more feasible solution rather than fusing them into one, especially...

Thanks for this example, i'll have a play around with it and see if i can't intuit how to extend this to my problem. > In your case, the container...

> Am I correct that the above example at least solves your problem of handling the data flow of input tensors to the module? Yes, as far as I can...

It seems like given the amount of operations in the [forward pass](https://github.com/m-lyon/dmri-pcconv/blob/33f758a941d32c25c76df30d28ce87fec506e79d/dmri_pcconv/core/model/layers/pcconv.py#L491), that refactoring these into a `Sequential` module wouldn't gain me much time/simplicity, over having to implement a function...

Thank you for your suggestions, I've set up a dev environment with `backpack` and am using the existing test suite to test the `jac_t_mat_prod` methods I derive. I've started from...

Hey Felix, I've written the following functionality to compute the `jac_t_mat_prod` for an arbitrary model for the first input ```python # backpack/core/derivates/model.py from typing import List, Tuple, Optional import torch...