fix inverse transform

api/app.py

@@ -79,6 +79,9 @@ def predict():
         predictions = model.predict(df)
         print("Predictions:", predictions)
 
+        predictions = scaler.inverse_transform(predictions.reshape(-1, 1))
+
+        print("Predictions conversed:", predictions)
         # Return predictions as JSON
         return jsonify({"predictions": predictions.tolist()})
 

src/predict.py

@@ -2,6 +2,17 @@
 import requests
 import hydra
 from data import read_features
+import zenml
+from hydra import compose, initialize
+import numpy as np
+
+
+with initialize(config_path="../configs", version_base=None):
+    cfg = compose(config_name="main")
+
+
+column_name = cfg.data.target_cols[0]
+scaler = zenml.load_artifact(f"{column_name}_scaler")
 
 
 # type: ignore
@@ -14,7 +25,6 @@ def predict(cfg=None):
 
     example = X.iloc[0, :]
     example_target = y[0]
-
     example = json.dumps({
         "inputs": example.to_dict()
     })
@@ -27,14 +37,18 @@ def predict(cfg=None):
         headers={"Content-Type": "application/json"},
     )
 
-    # print(predictions)
-    # predictions = scaler.inverse_transform(predictions.reshape(-1, 1))
-
-    # print("Predictions:", predictions)
 
-    print(response.json())
-    print("encoded target labels: ", example_target)
-    # print("target labels: ", list(cfg.data.labels)[example_target])
+
+    # take the value from {'predictions': [-0.467059463262558]}
+    prediction = response.json()["predictions"]
+    print("encoded prediction:", prediction)
+
+    prediction = np.array(prediction)
+    prediction = scaler.inverse_transform(prediction.reshape(-1, 1))
+    print("decoded prediction:", prediction)
+
+    print("encoded target labels:", example_target)
+    print("decoded target labels:", scaler.inverse_transform(np.array([example_target]).reshape(-1, 1)))
 
 
 if __name__ == "__main__":