github actions test

.github/workflows/test-external-models.yml

@@ -35,4 +35,5 @@ jobs:
 
     - name: Run
       run: |
-        chap evaluate --model-name https://github.com/sandvelab/chap_auto_ewars --dataset-name ISIMIP_dengue_harmonized --dataset-country brazil
+        #chap evaluate --model-name https://github.com/sandvelab/chap_auto_ewars --dataset-name ISIMIP_dengue_harmonized --dataset-country brazil
+        chap evaluate --model-name external_models/naive_python_model_with_mlproject_file_and_docker/ --dataset-name ISIMIP_dengue_harmonized --dataset-country brazil

climate_health/cli.py

@@ -10,6 +10,7 @@
 from cyclopts import App
 
 from climate_health.external.external_model import get_model_maybe_yaml
+from climate_health.external.mlflow import NoPredictionsError
 from climate_health.spatio_temporal_data.multi_country_dataset import MultiCountryDataSet
 from . import api
 from climate_health.dhis2_interface.ChapProgram import ChapPullPost
@@ -49,7 +50,11 @@ def evaluate(model_name: ModelType | str, dataset_name: DataSetType,  dataset_co
 
     model, model_name = get_model_maybe_yaml(model_name)
     model = model()
-    results = evaluate_model(model, dataset, prediction_length=prediction_length, n_test_sets=n_splits, report_filename=report_filename)
+    try:
+        results = evaluate_model(model, dataset, prediction_length=prediction_length, n_test_sets=n_splits, report_filename=report_filename)
+    except NoPredictionsError as e:
+        logging.error(f'No predictions were made: {e}')
+        return
     print(results)
 
 @app.command()

climate_health/external/mlflow.py

@@ -211,7 +211,7 @@ def predict(self, historic_data: DataSet, future_data: DataSet) -> DataSet:
         except pandas.errors.EmptyDataError:
             # todo: Probably deal with this in an other way, throw an exception istead
             logging.warning("No data returned from model (empty file from predictions)")
-            raise ValueError(f"No prediction data written")
+            raise NoPredictionsError(f"No prediction data written")
         result_class = SummaryStatistics if 'quantile_low' in df.columns else HealthData
         if self._location_mapping is not None:
             df['location'] = df['location'].apply(self._location_mapping.index_to_name)
@@ -220,3 +220,8 @@ def predict(self, historic_data: DataSet, future_data: DataSet) -> DataSet:
         mask = [start_time <= time_period.start_timestamp for time_period in time_periods]
         df = df[mask]
         return DataSet.from_pandas(df, Samples)
+
+
+
+class NoPredictionsError(Exception):
+    pass

external_models/naive_python_model_with_mlproject_file_and_docker/MLproject

@@ -0,0 +1,20 @@
+name: naive_python
+
+docker_env:
+  image: python:3.10
+
+#python_env: python_env.yaml
+
+entry_points:
+  train:
+    parameters:
+      train_data: str
+      model: str
+    command: "python train.py train {train_data} {model}"
+  predict:
+    parameters:
+      historic_data: str
+      future_data: str
+      model: str
+      out_file: str
+    command: "python  predict.py predict  {model} {historic_data} {future_data} {out_file}"

external_models/naive_python_model_with_mlproject_file_and_docker/predict.py

@@ -0,0 +1 @@
+print("predicting")

external_models/naive_python_model_with_mlproject_file_and_docker/train.py

@@ -0,0 +1,11 @@
+import sys
+
+train_data = sys.argv[1]
+model_name = sys.argv[2]
+
+
+with open(model_name, "w") as f:
+    f.write("test")
+
+
+