Updated scripts

Author: daenuprobst

data/az/az-2048-3-true-props.pkl.gz

@@ -52,7 +52,7 @@ def get_az_rxns(fold_idx: int = 0):
     for data, split in [(train, "train"), (valid, "valid"), (test, "test")]:
         X, mapping = DrfpEncoder.encode(
             data.smiles.to_numpy(),
-            n_folded_length=10240,
+            n_folded_length=2048,
             radius=3,
             rings=True,
             mapping=True,
@@ -66,18 +66,19 @@ def get_az_rxns(fold_idx: int = 0):
 
         y = data["yield"].to_numpy()
 
-        X_props = data.drop(
-            columns=[
-                "yield",
-                "reactant_smiles",
-                "solvent_smiles",
-                "base_smiles",
-                "product_smiles",
-                "id",
-            ]
-        ).to_numpy()
-
-        X = np.concatenate((X, X_props), axis=1)
+        # X_props = data.drop(
+        #     columns=[
+        #         "yield",
+        #         "reactant_smiles",
+        #         "solvent_smiles",
+        #         "base_smiles",
+        #         "product_smiles",
+        #         "id",
+        #         "smiles",
+        #     ]
+        # ).to_numpy()
+        #
+        # X = np.concatenate((X, X_props), axis=1)
 
         output_splits[split] = {
             "X": X,
@@ -88,8 +89,8 @@ def get_az_rxns(fold_idx: int = 0):
 
     output.append(output_splits)
 
-out_file_name = Path(az_path, f"az-10240-3-true-props.pkl")
-out_file_name_gz = Path(az_path, f"az-10240-3-true-props.pkl.gz")
+out_file_name = Path(az_path, f"az-2048-3-true.pkl")
+out_file_name_gz = Path(az_path, f"az-2048-3-true.pkl.gz")
 
 with open(out_file_name, "wb+") as f:
     pickle.dump(output, f)

data/az/az-2048-3-true.pkl.gz

@@ -4,6 +4,9 @@
 from statistics import stdev
 import numpy as np
 from xgboost import XGBRegressor
+from catboost import CatBoostRegressor
+from sklearn import svm
+from sklearn.ensemble import RandomForestRegressor
 from sklearn.metrics import r2_score, mean_absolute_error
 
 
@@ -21,7 +24,7 @@ def save_results(
 
 def predict_az():
     root_path = Path(__file__).resolve().parent
-    az_file_path = Path(root_path, "../../data/az/az-2048-3-true.pkl")
+    az_file_path = Path(root_path, "../../data/az/az-2048-3-true-props.pkl")
 
     data = pickle.load(open(az_file_path, "rb"))
 
@@ -59,9 +62,13 @@ def predict_az():
         )
 
         y_pred = model.predict(X_test, iteration_range=(0, model.best_iteration))
-        # y_pred[y_pred < 0.0] = 0.0
 
-        # save_results("az", split, sample_file, y_test, y_pred)
+        # X_train = np.concatenate((X_train, X_valid))
+        # y_train = np.concatenate((y_train, y_valid))
+        # model = RandomForestRegressor(n_estimators=1000, random_state=42)
+        # model.fit(X_train, y_train)
+        # y_pred = model.predict(X_test)
+
         r_squared = r2_score(y_test, y_pred)
         mae = mean_absolute_error(y_test, y_pred)
         print(f"Test {i + 1}", r_squared, mae / 100)

scripts/encoding/encode_az_reactions.py

@@ -37,7 +37,7 @@ def save_results(
 
 def predict_az():
     root_path = Path(__file__).resolve().parent
-    az_file_path = Path(root_path, "../../data/az/az-2048-3-true.pkl")
+    az_file_path = Path(root_path, "../../data/az/az-10240-3-true.pkl")
 
     data = pickle.load(open(az_file_path, "rb"))