Small refactor

[diegomarvid]

Generate

• Enable DataFrame for prediction, this allows to infer with real-time DataFrames without the need to have CSVs or Parquets. We can now do pipeline.predict(df).

Calculate Features

• Automatically ignore datetime_colums for training

Fit Model

• Merge with PredictStep
• Rename to ModelStep
• Rename model_params to model_parameters

Predict

• Delete step, now it's unnecessary

Model Registry

• Rename XGBoostModel to XGBoost
• Improve ModelClassNotFoundError error logging

README.md

• Update readme with new configuration examples

New configuration example

{
    "pipeline": {
        "name": "XGBoostTrainingPipeline",
        "description": "Training pipeline for XGBoost models.",
        "parameters": {
            "save_data_path": "test.pkl",
            "target": "target"
        },
        "steps": [
            {
                "step_type": "GenerateStep",
                "parameters": {
                    "train_path": "examples/ocf/data/dummy.csv",
                    "test_path": "examples/ocf/data/test.csv",
                    "predict_path": "examples/ocf/data/predict.csv"
                }
            },
            {
                "step_type": "TabularSplitStep",
                "parameters": {
                    "train_percentage": 0.95
                }
            },
            {
                "step_type": "CleanStep"
            },
            {
                "step_type": "CalculateFeaturesStep",
                "parameters": {
                    "datetime_columns": "date",
                    "features": [
                        "year",
                        "month",
                        "day"
                    ]
                }
            },
            {
                "step_type": "EncodeStep"
            },
            {
                "step_type": "ModelStep",
                "parameters": {
                    "model_class": "XGBoost",
                    "model_parameters": {
                        "n_estimators": 3,
                        "max_depth": 3
                    }
                }
            },
            {
                "step_type": "CalculateMetricsStep"
            }
        ]
    }
}

[ovejabu]

Nice refactor! LGTM