[good first issue - intermediate] Plot Materializer for Sklearn Plots

[skrawcz]

Is your feature request related to a problem? Please describe.
It is common for debugging scenarios in machine learning to produce plots for debugging purposes as a pipeline runs. You can do this with Hamilton, but we believe the ergonomics could be improved with materializers.

Describe the solution you'd like
Given the following Hamilton functions:

from sklearn.metrics import ConfusionMatrixDisplay, confusion_matrix

def y_pred(clf: Classifier, X_test: pd.Dataframe) -> pd.Series:
     return clf.predict(X_test)

def confusion_matrix(y_test: pd.Series,  y_pred: pd.Series) -> np.ndarray:
    return confusion_matrix(y_test, y_pred)

def cm_display(confusion_matrix: np.ndarray) -> ConfusionMatrixDisplay:
      return ConfusionMatrixDisplay(cm).plot()

We want to capture cm_display and save the produced plot somewhere.

One possible materializer API:

to.png(
   id="foo_bar_confusion_matrix",
   dependencies=["cm_display"]
   path="./foo_bar_confusion_matrix.png"
   height=12,
   withd=12, # some sort of kwargs...
)

That would handle taking the object and creating a PNG from it.

We should try to handle a few of the common plot types in sklearn.

Describe alternatives you've considered
You do this in the function -- but I think the materializer pattern helps abstract this.

[luahan77m]

Hi, I'm new to open source and want to try to solve some simple problems. Can you assign this issue to me? I would appreciate it!

[skrawcz]

@luahan77m sorry this issue has actually been done. I forgot to close it. Apologies.

If you want a simple task - #247 - is a good one to get started with. It will require you to become familiar with the process of contributing to open source and so I recommend that as a place to start.