chore: improve benchmarking abstraction

revision-scripts/mp_classification.py

@@ -20,8 +20,8 @@ def __len__(self):
         return self.txn.stat()['entries']
 
     def get(self, index):
-        id = f"{index}".encode("ascii")
-        return pickle.loads(self.txn.get(id))
+        id_ = f"{index}".encode("ascii")
+        return pickle.loads(self.txn.get(id_))
 
 def create_json_from_lmdb(lmdb_path, output_dir):
     dataset = Dataset(lmdb_path)

src/mattext/models/benchmark.py

@@ -138,7 +138,7 @@ def _record_predictions(self, task, fold, predictions, prediction_ids):
 
 class MatbenchmarkClassification(BaseBenchmark):
     def run_benchmarking(self, local_rank=None) -> None:
-        task = MatTextTask(task_name=self.task, is_classification=True)
+        task = self._initialize_task()
 
         for i, (exp_name, test_name) in enumerate(
             zip(self.exp_names, self.test_exp_names)
@@ -152,6 +152,9 @@ def run_benchmarking(self, local_rank=None) -> None:
 
         self._save_results(task)
 
+    def _initialize_task(self):
+        return MatTextTask(task_name=self.task, is_classification=True)
+
     def _get_finetuner(self, exp_cfg, local_rank, fold_name):
         return FinetuneClassificationModel(exp_cfg, local_rank, fold=fold_name)