Alexander Goscinski

We just would need to replace the one line in kernel computation for the case kernel_type `Full` to a numpy multiplication, right? This would still allow people to use the...

this is the current tutorials page  I agree that it is written the API, but we had a user who wasn't sure from the examples how to use sample...

> I was wondering how I can convert a model.json into a potential readable by Lammps. It can be directly read if you install a lammps version with our interface....

Suggestion for rename of the function: `_shallow_update(self, **hypers)` The private makes the user aware that it should be used with care and the hypers can be induced from the input...

We should use the hacky solution for a new `update_hyperparameters()` function

Ah sorry, I think I commented before finishing the whole comment. @cbenmahm had a solution, by basically creating a new Spherical(In|Co)variants object. "hacky" because we don't really update parameters, but...

Hello all, I think the output needs to be partially multiplied with the direction vectors so we get zero gradients for this frame. Firstly we need to consider the sum...

Small update, I tried if the same thing happens for a nonghost atoms (within the cell) and here one does not need to include the direction vector. So it is...

Hello, @Luthaf do you still get an error when you increase the cutoff as you suggested in your first response? I don't get it and I checked the gradient provider...

PR https://github.com/scikit-learn-contrib/scikit-matter/pull/201 moved examples to doctest for several of them, I ticket the boxes.