"""
MLflow experiment logging — three TODO blocks below record a training
run with MLflow.

The model and metric values in this script are synthetic. A trivial
DummyClassifier stands in for a trained model so that the MLflow
logging calls have a real sklearn estimator and deterministic numeric
metrics to persist. The purpose of the lab is to practise the MLflow
logging API, not to reason about model quality.

The three `# TODO` blocks inside the `mlflow.start_run()` context
are the only edits required.
"""
import numpy as np
import mlflow
import mlflow.sklearn
from sklearn.dummy import DummyClassifier

mlflow.set_tracking_uri("http://localhost:5000")

# Hyperparameters the run should record as MLflow parameters.
params = {"n_estimators": 100, "max_depth": 5, "random_state": 42}

# Synthetic "trained" model — a DummyClassifier fit on two deterministic
# rows so it has valid internal state for mlflow.sklearn.log_model to
# serialise. No real learning takes place.
X_fit = np.array([[0.0], [1.0]])
y_fit = np.array([0, 1])
model = DummyClassifier(strategy="most_frequent").fit(X_fit, y_fit)

# Synthetic evaluation scores. These are fixed values chosen for the
# lab — they are not computed from any real data.
accuracy = 0.92
f1 = 0.89

with mlflow.start_run():

    # TODO 1: log every entry in `params` as an MLflow parameter so that
    # n_estimators, max_depth, and random_state become searchable
    # parameters on this run.
    mlflow.log_params(params)
    


    # TODO 2: log `accuracy` and `f1` as MLflow metrics named
    # "accuracy" and "f1_score" respectively.
    mlflow.log_metric("accuracy", accuracy)
    mlflow.log_metric("f1_score", f1)

    # TODO 3: log the trained `model` as an MLflow sklearn model
    # artefact on this run.
    mlflow.sklearn.log_model(model, name="model")


    print(f"accuracy={accuracy}, f1_score={f1}")