fix

app.py

@@ -63,7 +63,7 @@ def init_space():
     #leaderboard_df = get_leaderboard_df(
         results_path=EVAL_RESULTS_PATH,
         requests_path=EVAL_REQUESTS_PATH,
-        #dynamic_path=DYNAMIC_INFO_FILE_PATH,
+        dynamic_path=DYNAMIC_INFO_FILE_PATH,
         cols=COLS,
         benchmark_cols=BENCHMARK_COLS
     )

src/leaderboard/read_evals.py

@@ -181,7 +181,7 @@ def get_request_file_for_model(requests_path, model_name, precision):
     return request_file
 
 
-def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResult]:
+def get_raw_eval_results(results_path: str, requests_path: str, dynamic_path: str) -> list[EvalResult]:
     """From the path of the results folder root, extract all needed info for results"""
     model_result_filepaths = []
 
@@ -199,11 +199,16 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
         for file in files:
             model_result_filepaths.append(os.path.join(root, file))
 
+    with open(dynamic_path) as f:
+        dynamic_data = json.load(f)
+
     eval_results = {}
     for model_result_filepath in model_result_filepaths:
         # Creation of result
         eval_result = EvalResult.init_from_json_file(model_result_filepath)
         eval_result.update_with_request_file(requests_path)
+        if eval_result.full_model in dynamic_data:
+                eval_result.update_with_dynamic_file_dict(dynamic_data[eval_result.full_model])
 
         # Store results of same eval together
         eval_name = eval_result.eval_name

src/populate.py

@@ -9,9 +9,9 @@ from src.leaderboard.read_evals import get_raw_eval_results
 from src.leaderboard.filter_models import filter_models_flags
 
 
-def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
+def get_leaderboard_df(results_path: str, requests_path: str, dynamic_path: str,cols: list, benchmark_cols: list) -> pd.DataFrame:
     """Creates a dataframe from all the individual experiment results"""
-    raw_data = get_raw_eval_results(results_path, requests_path)
+    raw_data = get_raw_eval_results(results_path, requests_path, dynamic_path)
     all_data_json = [v.to_dict() for v in raw_data]
     all_data_json.append(baseline_row)
     filter_models_flags(all_data_json)