disable progress bar in filtering and lower batch size to avoid segmentation fault

economy-watchers-survey-evaluation.py

@@ -1,6 +1,5 @@
 import datasets
 from datasets import load_dataset
-from datasets.utils.logging import disable_progress_bar
 
 # TODO: Add BibTeX citation
 # Find for instance the citation on arxiv or on the dataset repo/website
@@ -14,11 +13,11 @@ Dataset from the Economy Watchers Survey for use as evaluation tasks
 
 _HOMEPAGE = "https://github.com/retarfi/economy-watchers-survey"
 _LICENSE = "CC-BY 4.0"
-VERSION = datasets.Version("0.1.1")
+VERSION = datasets.Version("0.1.2")
 EWS_REPO_NAME = "retarfi/economy-watchers-survey"
 
 LABEL_REASON_OTHER: str = "それ以外"
-disable_progress_bar()
+DISABLED_BARS: bool = datasets.are_progress_bars_disabled()
 
 
 class EconomyWatchersSurveyConfig(datasets.BuilderConfig):
@@ -104,7 +103,7 @@ class EconomyWatchersSurveyDataset(datasets.GeneratorBasedBuilder):
         ]
 
     def _generate_examples(self, split: str):
-        batch_size = 2000
+        batch_size = 1000
         
         ds_info_current = load_dataset(
             EWS_REPO_NAME, "current", revision=self.config.ews_version, split=split
@@ -119,7 +118,11 @@ class EconomyWatchersSurveyDataset(datasets.GeneratorBasedBuilder):
                     split=f"{split}[{start_idx}:{end_idx}]"
                 )
                 
+                if not DISABLED_BARS:
+                    datasets.disable_progress_bars()
                 ds_current_batch = ds_current_batch.filter(lambda example: example["判断の理由"] is not None)
+                if not DISABLED_BARS:
+                    datasets.enable_progress_bars()
                 for example in ds_current_batch:
                     str_label = example["判断の理由"]
                     if str_label not in self.info.features["label"].names: