GSK-2690-return-empty-df-when-no-dataset (#81)

- return empty df when loading leaderboard dataset error (125e9f35a21808a3f302d39bbd05fd836930bf53)
- update the dataframe after every load from leaderboard dataset (4ecd715c49bb7e3a4daaf306c4c35a603990f253)
- clean log with logger.ingo (e842ae735c237b2a6049e74b5011e885f9f6cefe)


Co-authored-by: zcy <[email protected]>

app_leaderboard.py

@@ -2,6 +2,7 @@ import logging
 
 import datasets
 import gradio as gr
+import pandas as pd
 
 from fetch_utils import (check_dataset_and_get_config,
                          check_dataset_and_get_split)
@@ -24,7 +25,7 @@ def get_records_from_dataset_repo(dataset_id):
         logger.warning(
             f"Failed to load dataset {dataset_id} with config {dataset_config}: {e}"
         )
-        return None
+        return pd.DataFrame()
 
     
 def get_model_ids(ds):
@@ -73,10 +74,6 @@ def get_display_df(df):
         )
     return display_df
 
-def update_leaderboard_records():
-    logger.info("Updating leaderboard records")
-    leaderboard.records = get_records_from_dataset_repo(leaderboard.LEADERBOARD)
-
 def get_demo(leaderboard_tab):
     logger.info("Loading leaderboard records")
     leaderboard.records = get_records_from_dataset_repo(leaderboard.LEADERBOARD)
@@ -119,7 +116,15 @@ def get_demo(leaderboard_tab):
     with gr.Row():
         leaderboard_df = gr.DataFrame(display_df, datatype=types, interactive=False)
 
-    leaderboard_tab.select(fn=update_leaderboard_records)
+    def update_leaderboard_records(model_id, dataset_id, columns, task):
+        logger.info("Updating leaderboard records")
+        leaderboard.records = get_records_from_dataset_repo(leaderboard.LEADERBOARD)
+        return filter_table(model_id, dataset_id, columns, task)
+
+    leaderboard_tab.select(
+        fn=update_leaderboard_records, 
+        inputs=[model_select, dataset_select, columns_select, task_select], 
+        outputs=[leaderboard_df])
 
     @gr.on(
         triggers=[
@@ -132,6 +137,7 @@ def get_demo(leaderboard_tab):
         outputs=[leaderboard_df],
     )
     def filter_table(model_id, dataset_id, columns, task):
+        logger.info("Filtering leaderboard records")
         records = leaderboard.records
         # filter the table based on task
         df = records[(records["task"] == task)]