fix random splits names

ppb_affinity.py

@@ -39,23 +39,19 @@ class ppb_affinity(datasets.GeneratorBasedBuilder):
             ]
         elif self.config.name == "filtered_random":
             filepath = dl_manager.download_and_extract("filtered.csv")
-            # Read all rows to determine splits
             with open(filepath, encoding="utf-8") as f:
                 reader = csv.DictReader(f)
                 rows = list(reader)
             n_total = len(rows)
-            # Generate shuffled indices with fixed seed
             indices = list(range(n_total))
-            rng = random.Random(42)  # Fixed seed for reproducibility
+            rng = random.Random(42)
             rng.shuffle(indices)
-            # Calculate split sizes
             n_train = int(0.8 * n_total)
             n_val = int(0.1 * n_total)
-            n_test = n_total - n_train - n_val  # Handle remainder
-            # Split indices into ranges
+            n_test = n_total - n_train - n_val
             return [
                 datasets.SplitGenerator(
-                    name=datasets.NamedSplit("train_rand"),
+                    name=datasets.Split.TRAIN,
                     gen_kwargs={
                         "filepath": filepath,
                         "shuffled_indices": indices,
@@ -64,7 +60,7 @@ class ppb_affinity(datasets.GeneratorBasedBuilder):
                     },
                 ),
                 datasets.SplitGenerator(
-                    name=datasets.NamedSplit("validation_rand"),
+                    name=datasets.Split.VALIDATION,
                     gen_kwargs={
                         "filepath": filepath,
                         "shuffled_indices": indices,
@@ -73,7 +69,7 @@ class ppb_affinity(datasets.GeneratorBasedBuilder):
                     },
                 ),
                 datasets.SplitGenerator(
-                    name=datasets.NamedSplit("test_rand"),
+                    name=datasets.Split.TEST,
                     gen_kwargs={
                         "filepath": filepath,
                         "shuffled_indices": indices,
@@ -96,9 +92,8 @@ class ppb_affinity(datasets.GeneratorBasedBuilder):
                         del row["split"]
                         yield idx, row
             elif self.config.name == "filtered_random":
-                # Iterate over the range [split_start, split_end) in shuffled_indices
                 for global_idx in range(split_start, split_end):
                     original_idx = shuffled_indices[global_idx]
                     row = rows[original_idx]
-                    del row["split"]  # Remove original split column
-                    yield global_idx, row  # Key is global shuffled index
+                    del row["split"]
+                    yield global_idx, row