Fabian Zills
after the run method / when saving results we should os.chdir into the correct directory or use absolut paths
changing a parameter inside an experiment will also effect other experiments. This should not be the case.
We should acknowledge SPP and SimTech - the same for MDSuite.
ZnTrack Nodes are compatible along different packages. This only required a [Duck Typing](https://en.wikipedia.org/wiki/Duck_typing) approach. The Documentation should contain a part on this and highlight the importance of documenting the repsective...
DVC introduced a new `plots` key in the `dvc.yaml` (see https://github.com/iterative/dvc.org/issues/3763) This allows more complex definitions of plots which should be accessible through ZnTrack. Something like `dvc.plots(x=[val_loss, train_loss])` could be...
Feature request for e.g. benchmarking. Allow running custom code like `export CUDA_VISIBLE_DEVICES=-1` before starting the Python kernel as a custom argument to e.g. the `call` method.
It is difficult to write tests for #134 because they rely on class definitions that can be imported. So this is outsourced from the actual PR and should be solved...
If you try to make a Node a dependency that has no outputs, it should log a warning to reduce confusion.