eurobios-mews-labs / palma Goto Github PK

TypeError: TreeExplainer.init() got an unexpected keyword argument 'masker'

The ModelSelector API seems redundant with the Optimizer class :

Something like this would be more suitable and simple:

engine = FlamlOptimizer(**parameters)
engine.fit(project)

For consistency of naming we should have also

model = engine.best_model_
evaluator = ModelEvaluator(model=model) 

Rather than the confusing usage of estimator/model

Related issue40

It could be interessting to integrate a scoring visualisation for multi output classification

like :

$ pip install palma
$ python -c 'import palma'
[...]
ModuleNotFoundError: No module named 'deepchecks'

(Palma 2024.1.2)

You should be able to run the test suite even without the components optional dependencies.

Use importorskip to skip the test that require such an optional dependency.

To avoid giving different names to the same thing or the same name to different things

model = ModelEvaluation(estimator=ms.best_model_)

could be changed to something like

model_evaluation = ModelEvaluation(model=ms.best_model_)

Also for consistency, one might consider having either

• ModelSelection and ModelEvaluation
• or ModelSelector and ModelEvaluator

Hi Vincent,

Another message, just to share :
Linked to microsoft/FLAML#1193, I had to pip install "flaml[automl]"

Have a nice week,

Jérome

Hi Vincent,

I tried to install the palma package through :

pip install palma and pip install git+https://github.com/eurobios-mews-labs/palma.git

Each causes the error :

>>> import palma Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/dejaegherjer/monitoring_pylones/.venv/lib/python3.10/site-packages/palma/__init__.py", line 11, in <module> from palma.base.model import ModelEvaluation ModuleNotFoundError: No module named 'palma.base'

I succeed in installing it through :

git clone https://github.com/eurobios-mews-labs/palma.git pip install -e .

So no problem for me, just a message to share this problem, and maybe help if someone have the same problem.

Thank you for your job, cheers,

The current implementation is broken. I tried to fix it but I can't install auto-sklearn in the same environment as palma due to conflicting requirements with respect to the version of sklearn.