fix pipeline

Author: baturayo

.flake8

@@ -8,5 +8,5 @@ exclude =
     build,
     dist,
     tests
-max-line-length = 88
+max-line-length = 88 # black default
 ignore = D202,W503,E203 # conflicts with black

Makefile

@@ -15,7 +15,7 @@ NO_OF_REPORT_FILES := $(words $(filter-out reports/.gitkeep, $(SRC_FILES)))
 
 clean: ## clean artifacts
 	@echo ">>> cleaning files"
-	rm ./data/predictions/* ./data/transformed/* ./models/*.joblib
+	rm ./data/predictions/* ./data/transformed/* ./models/*.joblib || true
 
 generate-dataset: ## run ETL pipeline
 	@echo ">>> generating dataset"
@@ -29,9 +29,7 @@ serve: ## Serve model with a REST API using dploy-kickstart
 	@echo ">>> serving the trained model"
 	kickstart serve -e ml_skeleton_py/model/predict.py -l .
 
-run-pipeline: ## Run entire pipeline (clean artifacts -> generate-dataset -> train model -> serve model)
-	@echo ">>> running the pipeline"
-	clean clean generate-dataset train serve  ## clean artifacts generate dataset & train the model & serve
+run-pipeline: clean generate-dataset train serve  ## clean artifacts generate dataset & train the model & serve
 
 lint: ## flake8 linting and black code style
 	@echo ">>> black files"

scripts/train.py

@@ -1,6 +1,8 @@
 import argparse
+import os
 
 from ml_skeleton_py import model
+from ml_skeleton_py import settings as s
 
 
 def train() -> None:
@@ -19,7 +21,8 @@ def train() -> None:
         help="the serialized model name default lr " "referring to logistic regression",
     )
     args = parser.parse_args()
-    model.train(args.dataset, args.model_name)
+    transformed_data_dir = os.path.join(s.DATA_TRANSFORMED, args.dataset)
+    model.train(transformed_data_dir, s.MODEL_DIR, args.model_name)
 
 
 if __name__ == "__main__":