dfs

app.py

@@ -50,6 +50,7 @@ except Exception:
 
 
 LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
+LEADERBOARD_DF_N_CORRECT = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS, version="n_correct")
 
 (
     finished_eval_queue_df,
@@ -86,7 +87,7 @@ with demo:
             leaderboard = init_leaderboard(LEADERBOARD_DF)
  
         with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=1):
-            leaderboard = init_leaderboard(LEADERBOARD_DF)
+            leaderboard = init_leaderboard(LEADERBOARD_DF_N_CORRECT)
 
         with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table-n-correct", id=2):
             gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")

src/populate.py

@@ -8,9 +8,9 @@ from src.display.utils import AutoEvalColumn, EvalQueueColumn
 from src.leaderboard.read_evals import get_raw_eval_results
 
 
-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, cols: list, benchmark_cols: list, version="1_correct") -> 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+"/"+version, requests_path)
     all_data_json = [v.to_dict() for v in raw_data]
 
     print(all_data_json)