update

jester_rating.py

@@ -1,7 +1,7 @@
 import pandas as pd
 import datasets
-from pandas._libs.algos import groupsort_indexer
-from sklearn.model_selection import GroupShuffleSplit
+from sklearn.model_selection import train_test_split
+import numpy as np
 
 _CITATION = "N/A"
 _DESCRIPTION = "N/A" 
@@ -58,33 +58,38 @@ class JesterEmbedding(datasets.GeneratorBasedBuilder):
         ]
 
     def _generate_examples(self, filepath, split):
-        ratings_df = pd.read_parquet(filepath)
+        df = pd.read_parquet(filepath)
 
-        for x in GroupShuffleSplit(n_splits=1, test_size=0.2, random_state=42).split(ratings_df, groups=ratings_df.user_id):
-            train_idx, test_idx = x
+        rng = np.random.RandomState(42)
         
-        _train_ratings = ratings_df.iloc[train_idx]
-        test_ratings_df = ratings_df.iloc[test_idx]
+        train_df, test_df, val_df = [], [], []
         
-        for x in GroupShuffleSplit(n_splits=1, test_size=0.2, random_state=7).split(_train_ratings, groups=_train_ratings.user_id):
-            train_idx, val_idx = x
+        for user_id in df.user_id.unique():
+            if len(df[df.user_id == user_id]) < 3:
+                continue
+            _train_df, _test_df = train_test_split(df[df.user_id == user_id], test_size=0.2, random_state=rng)
+            _train_df, _val_df = train_test_split(_train_df, test_size=0.2, random_state=rng)
         
-        train_ratings_df = _train_ratings.iloc[train_idx]
-        val_ratings_df = _train_ratings.iloc[val_idx]
+            train_df.append(_train_df)
+            val_df.append(_val_df)
+            test_df.append(_test_df)
         
+        train_df = pd.concat(train_df)
+        test_df = pd.concat(test_df)
+        val_df = pd.concat(val_df)
 
         if split == "train":
-            for _id, row in train_ratings_df.iterrows():
+            for _id, row in train_df.iterrows():
                 user_id, item_id, rating = row
                 user_id, item_id = int(user_id), int(item_id)
                 yield _id, {"user_id": user_id, "item_id": item_id, "rating": rating}
         elif split == "test":
-            for _id, row in test_ratings_df.iterrows():
+            for _id, row in test_df.iterrows():
                 user_id, item_id, rating = row
                 user_id, item_id = int(user_id), int(item_id)
                 yield _id, {"user_id": user_id, "item_id": item_id, "rating": rating}
         else:
-            for _id, row in val_ratings_df.iterrows():
+            for _id, row in val_df.iterrows():
                 user_id, item_id, rating = row
                 user_id, item_id = int(user_id), int(item_id)
                 yield _id, {"user_id": user_id, "item_id": item_id, "rating": rating}