Multi-class Classification examples #28
Conversation
Signed-off-by: mshtelma <[email protected]>
…dded hyperparameter tuning to the multi-class example
| @@ -1,4 +1,4 @@ | |||
| mlflow>=2.0 | |||
There was a problem hiding this comment.
Is there a reason we removed the mlflow requirement?
There was a problem hiding this comment.
Thanks, good catch! Added it back.
| ipykernel>=6.12 | ||
| ipython>=7.32 | ||
| flaml |
There was a problem hiding this comment.
Do we need flaml overall? Or do you think it is a good idea to add it in the notebook or create a sub requirement for this? Thoughts?
There was a problem hiding this comment.
That's a good point. I think there are other examples that use FLAML by default (at least I saw some PRs), so this might be a good idea. This is not a big dependency. I am happy to move it inside the multi-class folder as well.
There was a problem hiding this comment.
I recall flaml is problematic to install with lightgbm installation difficult. Has this been fixed? If not, prefer to not including it. It's not used in this example anyway.
mshtelma
changed the title
| @@ -0,0 +1,26 @@ | |||
| # Binary classification: Is this bottle of wine red or white? | |||
| This is the root directory for an example project for the | |||
| [MLflow Classification Recipe](https://mlflow.org/docs/latest/recipes.html#classification-recipe). | |||
There was a problem hiding this comment.
[MLflow Classification Recipe] --> [MLflow Multi Class Classification Recipe]
What do you think?
| # COMMAND ---------- | ||
|
|
||
| # MAGIC %pip install -r ../../requirements.txt | ||
| # MAGIC %pip install git+https://github.com/mshtelma/mlflow.git@multiclassclassification |
There was a problem hiding this comment.
Remove this once MLFlow is released.
There was a problem hiding this comment.
We just released MLflow 2.1.0 so we should be good to remove this now :)
| # COMMAND ---------- | ||
|
|
||
| r.run("split") |
There was a problem hiding this comment.
Can we do a bit of EDA here before proceeding to split?
|
|
||
| # COMMAND ---------- | ||
|
|
||
| trained_model = r.get_artifact("model") |
There was a problem hiding this comment.
Can we use the model to make a prediction on one hand-generated input example here?
| # For different options please read: https://github.com/mlflow/recipes-classification-template#ingest-step | ||
| using: csv | ||
| loader_method: load_file_as_dataframe | ||
| location: "./data/iris.csv" |
There was a problem hiding this comment.
Can this be an open access Databricks Delta table? Does it exist?
| @@ -0,0 +1,31 @@ | |||
| experiment: | |||
| name: "sklearn_classification_experiment" | |||
There was a problem hiding this comment.
Update the experiment name to not collide with binary classification example
| INGEST_SCORING_CONFIG: | ||
| # For different options please read: https://github.com/mlflow/recipes-classification-template#batch-scoring | ||
| using: csv | ||
| location: "./data/iris.csv" |
There was a problem hiding this comment.
It'd be nice to use a different dataset for scoring, if such dataset exists. Or you can create one. See https://github.com/mlflow/recipes-examples/blob/main/classification/profiles/local.yaml#L26 for example.
| import pandas | ||
|
|
||
| df = pandas.read_csv(file_path, sep=",") | ||
| df["class"] = df["class"].astype("category").cat.codes |
There was a problem hiding this comment.
What does this line do? What happens if there's no such conversion?
| transformers. | ||
| """ | ||
|
|
||
| return None |
There was a problem hiding this comment.
Add a comment indicating this will result in an identical transformer.
| :return: A Series indicating whether each row should be filtered | ||
| """ | ||
|
|
||
| return Series(True, index=dataset.index) |
There was a problem hiding this comment.
Add a comment to indicate this doesn't filter out anything.
| @@ -0,0 +1,10 @@ | |||
| from steps.transform import transformer_fn | |||
There was a problem hiding this comment.
Why introduce an empty file multi-class-classification/tests/train_test.py above?
| ipykernel>=6.12 | ||
| ipython>=7.32 | ||
| flaml |
There was a problem hiding this comment.
I recall flaml is problematic to install with lightgbm installation difficult. Has this been fixed? If not, prefer to not including it. It's not used in this example anyway.
No description provided.