gridrdf functionality

.pre-commit-config.yaml

@@ -6,6 +6,6 @@ repos:
 #     -   id: end-of-file-fixer
 #     -   id: trailing-whitespace
 -   repo: https://github.com/psf/black
-    rev: 19.3b0
+    rev: 23.3.0
     hooks:
     -   id: black

examples/crabnet_performance.py

@@ -4,6 +4,7 @@
 # TODO: incorporate CrabNet uncertainty into search
 """
 from crabnet.crabnet_ import CrabNet
+import pandas as pd
 
 # %% imports
 from tqdm import tqdm
@@ -82,7 +83,7 @@
         idx = perf_val_df.pred.idxmax()
         # idx = np.where(val_pred == max(val_pred))[0][0]
         move_row = perf_val_df.loc[idx]
-        perf_train_df.append(move_row)
+        perf_train_df = pd.concat((perf_train_df, move_row))
         perf_val_df = perf_val_df.drop(index=idx)
         next_experiments.append(move_row.to_dict())
     experiment = ad_experiments_metrics(

mat_discover/adaptive_design.py

@@ -150,7 +150,7 @@ def suggest_next_experiment(
         # append compound to train, remove from val, and reset indices
         # https://stackoverflow.com/a/12204428/13697228
         move_row = self.val_df[self.val_df.index == next_index]
-        self.train_df = self.train_df.append(move_row)
+        self.train_df = pd.concat((self.train_df, move_row))
         self.val_df = self.val_df[self.val_df.index != next_index]
         # self.val_df = self.val_df.drop(index=next_index)
 

mat_discover/mat_discover_.py

@@ -813,8 +813,8 @@ def predict(
                 f"self.val_rad_neigh_avg` and `self.val_k_neigh_avg` are being assigned the same values as `val_dens` for compatibility reasons since a non-DiSCoVeR novelty learner was specified: {self.novelty_learner}."
             )
             # composition-based featurization
-            X_train: Union[pd.DataFrame, np.ndarray, List] = []
-            X_val: Union[pd.DataFrame, np.ndarray, List] = []
+            X_train: Union[np.ndarray, List] = []
+            X_val: Union[np.ndarray, List] = []
             assert self.train_inputs is not None
             if self.novelty_prop == "mod_petti":
                 assert isinstance(X_train, list)

mat_discover/utils/extraordinary.py

@@ -17,7 +17,7 @@ def extraordinary_split(
 ):
     # set aside high-performing candidates
     if val_df is not None:
-        train_val_df = train_df.append(val_df)
+        train_val_df = pd.concat((train_df, val_df))
     else:
         train_val_df = train_df
 
@@ -31,7 +31,7 @@ def extraordinary_split(
     train_df, val_df = train_test_split(
         train_val_df, train_size=train_size, random_state=random_state
     )
-    val_df = val_df.append(extraordinary_df)
+    val_df = pd.concat((val_df, extraordinary_df))
 
     return train_df, val_df, extraordinary_thresh
 

pyproject.toml

@@ -33,6 +33,7 @@ dependencies = [
     "umap-learn",
     "dill",
     "crabnet >=2.0.5,<3.0.0",
+    "torch <2",
     "chem_wasserstein >=1.0.8,<2.0.0",
     "composition_based_feature_vector",
     "matbench_genmetrics >= 0.6.1",