Merge branch 'main' into operators/automlx_24.2

Author: ahosler

.github/workflows/run-forecast-unit-tests.yml

@@ -31,6 +31,10 @@ jobs:
 
     steps:
       - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ github.event.pull_request.head.sha }}
+
 
       - uses: actions/setup-python@v5
         with:

THIRD_PARTY_LICENSES.txt

@@ -447,6 +447,12 @@ python-fire
 * Source code: https://github.com/google/python-fire
 * Project home: https://github.com/google/python-fire
 
+mlforecast
+* Copyright 2024 Nixtla
+* License: Apache License 2.0
+* Source code: https://github.com/Nixtla/mlforecast
+* Project home: https://github.com/Nixtla/mlforecast
+
 =======
 =============================== Licenses ===============================
 ------------------------------------------------------------------------

ads/jobs/builders/infrastructure/dsc_job.py

@@ -9,6 +9,7 @@
 import inspect
 import logging
 import os
+import re
 import time
 import traceback
 import uuid
@@ -375,12 +376,13 @@ def delete(self, force_delete: bool = False) -> DSCJob:
         """
         runs = self.run_list()
         for run in runs:
-            if run.lifecycle_state in [
-                DataScienceJobRun.LIFECYCLE_STATE_ACCEPTED,
-                DataScienceJobRun.LIFECYCLE_STATE_IN_PROGRESS,
-                DataScienceJobRun.LIFECYCLE_STATE_NEEDS_ATTENTION,
-            ]:
-                run.cancel(wait_for_completion=True)
+            if force_delete:
+                if run.lifecycle_state in [
+                    DataScienceJobRun.LIFECYCLE_STATE_ACCEPTED,
+                    DataScienceJobRun.LIFECYCLE_STATE_IN_PROGRESS,
+                    DataScienceJobRun.LIFECYCLE_STATE_NEEDS_ATTENTION,
+                ]:
+                    run.cancel(wait_for_completion=True)
             run.delete()
         self.client.delete_job(self.id)
         return self
@@ -582,6 +584,25 @@ def logging(self) -> OCILog:
             id=self.log_id, log_group_id=self.log_details.log_group_id, **auth
         )
 
+    @property
+    def exit_code(self):
+        """The exit code of the job run from the lifecycle details.
+        Note that,
+        None will be returned if the job run is not finished or failed without exit code.
+        0 will be returned if job run succeeded.
+        """
+        if self.lifecycle_state == self.LIFECYCLE_STATE_SUCCEEDED:
+            return 0
+        if not self.lifecycle_details:
+            return None
+        match = re.search(r"exit code (\d+)", self.lifecycle_details)
+        if not match:
+            return None
+        try:
+            return int(match.group(1))
+        except Exception:
+            return None
+
     @staticmethod
     def _format_log(message: str, date_time: datetime.datetime) -> dict:
         """Formats a message as log record with datetime.
@@ -655,6 +676,22 @@ def _check_and_print_status(self, prev_status) -> str:
             print(f"{timestamp} - {status}")
         return status
 
+    def wait(self, interval: float = SLEEP_INTERVAL):
+        """Waits for the job run until if finishes.
+
+        Parameters
+        ----------
+        interval : float
+            Time interval in seconds between each request to update the logs.
+            Defaults to 3 (seconds).
+
+        """
+        self.sync()
+        while self.status not in self.TERMINAL_STATES:
+            time.sleep(interval)
+            self.sync()
+        return self
+
     def watch(
         self,
         interval: float = SLEEP_INTERVAL,
@@ -830,6 +867,12 @@ def download(self, to_dir):
         self.job.download(to_dir)
         return self
 
+    def delete(self, force_delete: bool = False):
+        if force_delete:
+            self.cancel(wait_for_completion=True)
+        super().delete()
+        return
+
 
 # This is for backward compatibility
 DSCJobRun = DataScienceJobRun

ads/opctl/operator/lowcode/forecast/const.py

@@ -14,6 +14,7 @@ class SupportedModels(str, metaclass=ExtendedEnumMeta):
     Prophet = "prophet"
     Arima = "arima"
     NeuralProphet = "neuralprophet"
+    MLForecast = "mlforecast"
     AutoMLX = "automlx"
     AutoTS = "autots"
     Auto = "auto"

ads/opctl/operator/lowcode/forecast/environment.yaml

@@ -8,6 +8,7 @@ dependencies:
       - oracle-ads>=2.9.0
       - prophet
       - neuralprophet
+      - mlforecast
       - pmdarima
       - statsmodels
       - report-creator

ads/opctl/operator/lowcode/forecast/model/factory.py

@@ -12,6 +12,7 @@
 from .base_model import ForecastOperatorBaseModel
 from .neuralprophet import NeuralProphetOperatorModel
 from .prophet import ProphetOperatorModel
+from .ml_forecast import MLForecastOperatorModel
 from ..utils import select_auto_model
 from .forecast_datasets import ForecastDatasets
 
@@ -32,6 +33,7 @@ class ForecastOperatorModelFactory:
         SupportedModels.Prophet: ProphetOperatorModel,
         SupportedModels.Arima: ArimaOperatorModel,
         SupportedModels.NeuralProphet: NeuralProphetOperatorModel,
+        SupportedModels.MLForecast: MLForecastOperatorModel,
         SupportedModels.AutoMLX: AutoMLXOperatorModel,
         SupportedModels.AutoTS: AutoTSOperatorModel
     }

ads/opctl/operator/lowcode/forecast/model/forecast_datasets.py

@@ -86,7 +86,10 @@ def create_horizon(self, spec, historical_data):
             pd.date_range(
                 start=historical_data.get_max_time(),
                 periods=spec.horizon + 1,
-                freq=historical_data.freq,
+                freq=historical_data.freq
+                or pd.infer_freq(
+                    historical_data.data.reset_index()[spec.datetime_column.name][-5:]
+                ),
             ),
             name=spec.datetime_column.name,
         )
@@ -135,6 +138,7 @@ def __init__(self, config: ForecastOperatorConfig):
 
         self._horizon = config.spec.horizon
         self._datetime_column_name = config.spec.datetime_column.name
+        self._target_col = config.spec.target_column
         self._load_data(config.spec)
 
     def _load_data(self, spec):
@@ -158,6 +162,16 @@ def get_all_data_long(self, include_horizon=True):
             on=[self._datetime_column_name, ForecastOutputColumns.SERIES],
         ).reset_index()
 
+    def get_all_data_long_forecast_horizon(self):
+        """Returns all data in long format for the forecast horizon."""
+        test_data = pd.merge(
+            self.historical_data.data,
+            self.additional_data.data,
+            how="outer",
+            on=[self._datetime_column_name, ForecastOutputColumns.SERIES],
+        ).reset_index()
+        return test_data[test_data[self._target_col].isnull()].reset_index(drop=True)
+
     def get_data_multi_indexed(self):
         return pd.concat(
             [