Dataset Version

src/display/utils.py

@@ -34,6 +34,7 @@ auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sh
 
 
 auto_eval_column_dict.append(["output_format", ColumnContent, ColumnContent("Output Format", "str", False, False)])
+auto_eval_column_dict.append(["dataset_version", ColumnContent, ColumnContent("Task Version", "str", False, False)])
 
 # We use make dataclass to dynamically fill the scores from Tasks
 AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True)

src/leaderboard/read_evals.py

@@ -32,6 +32,7 @@ class EvalResult:
     date: str = "" # submission date of request file
     still_on_hub: bool = False
     output_format: str = "?" # output format of the model
+    dataset_version: str = "?" # version of the dataset used
 
     @classmethod
     def init_from_json_file(self, json_filepath):
@@ -68,6 +69,7 @@ class EvalResult:
                 architecture = ";".join(architectures)
 
         output_format = config.get("output_format", "?")
+        dataset_version = config.get("dataset_version", "?")
 
         # Extract results available in this file (some results are split in several files)
         results = {}
@@ -93,6 +95,7 @@ class EvalResult:
             still_on_hub=still_on_hub,
             architecture=architecture,
             output_format=output_format,
+            dataset_version=dataset_version,
         )
 
     def update_with_request_file(self, requests_path):
@@ -121,6 +124,7 @@ class EvalResult:
             AutoEvalColumn.revision.name: self.revision,
             AutoEvalColumn.average.name: average,
             AutoEvalColumn.output_format.name: self.output_format,
+            AutoEvalColumn.dataset_version.name: self.dataset_version,
         }
 
         for task in Tasks: