Fabian Zills

Something like `metrics = {rmse: {key: value}}` should be `metrics.rmse.key`

Improve `load` method

1

Currently there are (at least) two things missing: ```python node = SomeNode(name="abc") node.load() # this will return SomeNode.load() and NOT SomeNode.load("abc") ``` in place is not possible: ```python node.load() #...

If the `dvc.params` is not of a supported type it will be treated like a depdendency. Maybe make an Issue on the DVC github.

In addition to pandas dataframes lists and arrays should also be possible.

Using `os.chdir` in the run method leads to an AttributeError. Can potentially be fixed by using absolute paths. Maybe as backup?

When running a lot of simulations support for MLFlow would be nice

You need something like: ``` Node().write_graph(queue=True) # or zntrack.queue([Node(), Node(), ...]) ```

Assume you change your Node from ```python class MyNode(Node): input = zn.params() data = zn.outs() ``` to ```python class MyNode(Node): seed = zn.params() input = zn.params() data = zn.outs() ```...

Simulation data post processing

2

- [x] Add category **Category details:** In atomistic ML we face new challenges for post processing data. I would suggest this category to include existing tools from e.g. classical MD...

## Description The link to the plumed documentation went 404. This should fix this. ## Status - [x] Ready to go