fix: Use native xgb.Booster to fix Py3.12 crash (keeps .ubj format)

pyproject.toml

@@ -8,7 +8,7 @@ dynamic = ["version"]
 description = "Open Source Solar Forecasting for a Site"
 authors = [{ name = "Peter Dudfield", email = "info@openclimatefix.org" }]
 readme = "README.md"
-requires-python = ">=3.11"
+requires-python = ">=3.11,<3.13"
 license = { text = "MIT" }
 
 dependencies = [
@@ -27,7 +27,7 @@ dependencies = [
     "pydantic_settings",
     "httpx",
     "sentry_sdk",
-    "huggingface_hub==0.17.3"
+    "huggingface_hub==0.21.4"
 ]
 
 [project.urls]

quartz_solar_forecast/forecasts/v2.py

@@ -7,6 +7,7 @@
 import pandas as pd
 from huggingface_hub import hf_hub_download
 from xgboost.sklearn import XGBRegressor
+import xgboost as xgb
 
 import quartz_solar_forecast
 from quartz_solar_forecast.weather import WeatherService
@@ -91,7 +92,7 @@ def _decompress_zipfile(self, filename: str) -> None:
 
     def load_model(
         self,
-        model_file: str = constants.MODEL_FILE,
+        model_file: str = "model_10_202405.ubj",
         repo_id: str = "openclimatefix/open-source-quartz-solar-forecast",
         file_path: str = "models/v2/model_10_202405.ubj.zip",
     ) -> XGBRegressor:
@@ -115,21 +116,13 @@ def load_model(
         XGBRegressor
             The loaded XGBoost model ready for making predictions.
         """
-        # Use the project directory
-        zipfile_model = os.path.join(self.download_dir, model_file + ".zip")
-
-        if not os.path.isfile(zipfile_model):
-            logger.info("Downloading model...")
-            zipfile_model = self._download_model(model_file + ".zip", repo_id, file_path)
 
         model_path = os.path.join(self.download_dir, model_file)
-        if not os.path.isfile(model_path):
-            logger.info("Preparing model...")
-            self._decompress_zipfile(zipfile_model)
 
-        logger.info("Loading model...")
-        loaded_model = XGBRegressor()
+        logger.info(f"Loading Raw Booster from {model_path}...")
+        loaded_model = xgb.Booster()
         loaded_model.load_model(model_path)
+        loaded_model._estimator_type = "regressor"
         self.model = loaded_model
         return loaded_model
 
@@ -260,14 +253,13 @@ def predict_power_output(
         """
 
         data = self.get_data(latitude, longitude, start_date, kwp, orientation, tilt)
-        # if data is not None:
         cleaned_data = self.clean(data)
-        predictions = self.model.predict(cleaned_data.drop(columns=[self.DATE_COLUMN]))
+        X = cleaned_data.drop(columns=[self.DATE_COLUMN])
+        dmatrix = xgb.DMatrix(X)
+        predictions = self.model.predict(dmatrix)
         predictions_df = pd.DataFrame(predictions, columns=["prediction"])
         final_data = cleaned_data.join(predictions_df)
-        # set night predictions to 0
         final_data.loc[final_data["is_day"] == 0, "prediction"] = 0
-        # set negative output to 0
         final_data.loc[final_data["prediction"] < 0, "prediction"] = 0
         df = final_data[[self.DATE_COLUMN, "prediction"]]
         df = df.rename(columns={"prediction": "power_kw"})

quartz_solar_forecast/utils/sentry_logging.py

@@ -1,10 +1,10 @@
-"""Log usage of this package to Sentry"""
-
 import importlib.metadata
 import os
 
 import sentry_sdk
 
+from sentry_sdk.integrations.huggingface_hub import HuggingfaceHubIntegration
+
 from quartz_solar_forecast.pydantic_models import PVSite
 
 version = importlib.metadata.version("quartz_solar_forecast")
@@ -47,4 +47,4 @@ def write_sentry(params):
         sentry_sdk.set_tag("version", version)
 
     except Exception as _:  # noqa
-        pass
+        pass

quartz_solar_forecast/weather/open_meteo.py

@@ -98,9 +98,8 @@
         except ValueError as e:
             raise ValueError(
                 f"Invalid date format. Please use YYYY-MM-DD format. Error: {str(e)}"
-            ) from e
-
+
         if not (end_datetime > start_datetime):
             raise ValueError(
                 f"Invalid date range. End date ({end_date}) must be greater than "
                 f"start date ({start_date})."
@@ -166,6 +165,7 @@
         except requests.exceptions.Timeout as e:
             raise TimeoutError(f"Request to OpenMeteo API timed out. URl - {url}") from e
 
+        # Process the hourly data
         hourly = response[0].Hourly()
         hourly_data = {
             "time": pd.date_range(