ZnTrack v0.8

[PythonFZ]

• [ ] check what can be passed to zntrack.deps (after #825) Node, Con, dataclass
• [ ] notebook support using https://github.com/zincware/znipy ?
• [x] dataclass deps need to save the module to params (as well? or is this enough) to ensure two dataclasses with the same parameters but different modules will trigger runs. (test!)
• [ ] fix __repr__ issues. A log.info(node) should never be the cause of an exception.
• [ ] test _post_init_
• [ ] test setting an output / parameter twice (https://github.com/zincware/znfields/issues/4)
• [x] test dependency to a node that has no outputs
• [x] restructure zntrack on top of dataclasses
• [ ] write as znflow deployment
• [ ] proj.run should also build the project.
• [x] plugins: DVC, AIM, MLFlow, wandb
• [ ] have an option, to append to the dvc.yaml and not overwrite everything, when adding non-ZnTrack stages.
• [ ] build project with deps but without params - "no such file as params.yaml"
• [ ] test custom plugin outputs
• [ ] git ignore dvc tracked files (not using dvc stage add this might not be done?)
• [x] external nodes, e.g. md thermostats
• [x] test with project.group and with projget.group inside with project
• [ ] zntrack.outs() with np.ndarray is not properly deserialized when loading
• [ ] log automatic_node_names exception when passed. (move the graph to project.graph instead of inheritance)
• [ ] xx_path support tuple[Path, ...]
• [x] https://github.com/zincware/ZnTrack/blob/4b260fdc492e5fdc5fa38624ca866ecbac9a7bb7/zntrack/converter.py#L108-L116
• [ ] need tests for all features also using from_rev with rev and remote.
• [ ] use lazy loading for plugins to skip the module
• [ ] zntrack.plots_path() (have an additional kwarg to specify, that the file is already tracked via outs and this is just a plots configuration).
• [x] plots only work if x and y are given. Need to have an index on the dataframe for x
• [ ] for external, convert the external node in a node on the graph which imports the data e.g. via --method node.state.import.<attribute> or something to import the data not while building the graph but while running the node? This would alter the dvc.yaml though.
• [ ] aim finalize yields TypeError: Unhandled non-native value 0of type<class 'numpy.int64'>. ?

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 95.60000% with 44 lines in your changes missing coverage. Please review.

Project coverage is 94.83%. Comparing base (d602167) to head (25c9088).

Files with missing lines	Patch %	Lines
zntrack/utils/misc.py	81.69%	13 Missing :warning:
zntrack/converter.py	96.33%	4 Missing :warning:
zntrack/group.py	92.98%	4 Missing :warning:
zntrack/plugins/base.py	90.47%	4 Missing :warning:
zntrack/cli.py	80.00%	3 Missing :warning:
zntrack/node.py	96.00%	3 Missing :warning:
zntrack/state.py	97.08%	3 Missing :warning:
zntrack/utils/import_handler.py	85.71%	3 Missing :warning:
zntrack/project.py	97.61%	2 Missing :warning:
zntrack/utils/node_wd.py	93.33%	2 Missing :warning:
... and 3 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #812      +/-   ##
==========================================
+ Coverage   91.13%   94.83%   +3.69%     
==========================================
  Files          26       21       -5     
  Lines        2075     1103     -972     
  Branches      985      549     -436     
==========================================
- Hits         1891     1046     -845     
+ Misses        184       57     -127     

Flag	Coverage Δ
fulltest	94.28% <95.00%> (+7.34%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[PythonFZ]

For mlflow base the parent run on the base commit, like dvc experiments do it? What if multiple experiments are based on one commit?