COLS for multimodal track

app.py

@@ -17,6 +17,7 @@ from src.display.utils import (
     BENCHMARK_COLS,
     BENCHMARK_COLS_MULTIMODAL,
     COLS,
+    COLS_MULTIMODAL,
     EVAL_COLS,
     EVAL_TYPES,
     AutoEvalColumn,
@@ -48,7 +49,7 @@ except Exception:
 
 
 LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
-LEADERBOARD_DF_MULTIMODAL = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS_MULTIMODAL)
+LEADERBOARD_DF_MULTIMODAL = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS_MULTIMODAL, BENCHMARK_COLS_MULTIMODAL)
 
 (
     finished_eval_queue_df,

src/display/utils.py

@@ -22,18 +22,25 @@ class ColumnContent:
 
 ## Leaderboard columns
 auto_eval_column_dict = []
+auto_eval_column_dict_multimodal = []
 # Init
 auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
 #Scores
 auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Average ⬆️", "number", True)])
+auto_eval_column_dict_multimodal = auto_eval_column_dict
 for task in Tasks:
     auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
+for task in TasksMultimodal:
+    auto_eval_column_dict_multimodal.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
 # Model information
 auto_eval_column_dict.append(["still_on_hub", ColumnContent, ColumnContent("Available on the hub", "bool", False)])
 auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
+auto_eval_column_dict_multimodal.append(["still_on_hub", ColumnContent, ColumnContent("Available on the hub", "bool", False)])
+auto_eval_column_dict_multimodal.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
 
 # We use make dataclass to dynamically fill the scores from Tasks
 AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True)
+AutoEvalColumnMultimodal = make_dataclass("AutoEvalColumnMultimodal", auto_eval_column_dict_multimodal, frozen=True)
 
 ## For the queue columns in the submission tab
 @dataclass(frozen=True)
@@ -53,6 +60,7 @@ class ModelDetails:
 
 # Column selection
 COLS = [c.name for c in fields(AutoEvalColumn) if not c.hidden]
+COLS_MULTIMODAL = [c.name for c in fields(AutoEvalColumnMultimodal) if not c.hidden]
 
 EVAL_COLS = [c.name for c in fields(EvalQueueColumn)]
 EVAL_TYPES = [c.type for c in fields(EvalQueueColumn)]

src/populate.py

@@ -23,7 +23,7 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
     df = pd.DataFrame.from_records(all_data_json)
     print(df)
     df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
-    df = df[benchmark_cols].round(decimals=1)
+    df = df[cols].round(decimals=1)
 
     # filter out if any of the benchmarks have not been produced
     df = df[has_no_nan_values(df, benchmark_cols)]