Update app.py

app.py

@@ -290,73 +290,6 @@ def _round_df3(df: pd.DataFrame) -> pd.DataFrame:
             df[c] = df[c].round(3)
     return df
 
-# def predict_files(
-#     files: List[str],
-#     model1_name: str,
-#     model2_name: str,
-#     ckpt_map_json: str,
-#     compact: bool
-# ) -> pd.DataFrame:
-#     if not model1_name:
-#         raise gr.Error("Please select Model 1.")
-
-#     ckpt_map = json.loads(ckpt_map_json)
-
-#     repo_file1 = ckpt_map[model1_name]
-#     m1 = load_model_from_hub(model1_name, repo_file1)
-#     arch1 = arch_from_filename(repo_file1)
-
-#     m2, arch2 = None, None
-#     if model2_name:
-#         repo_file2 = ckpt_map[model2_name]
-#         m2 = load_model_from_hub(model2_name, repo_file2)
-#         arch2 = arch_from_filename(repo_file2)
-
-#     rows = []
-#     for f in files:
-#         try:
-#             img, base = _open_image(f)
-#         except Exception:
-#             continue
-
-#         L1 = logits_for_model(m1, arch1, img)
-#         L2 = logits_for_model(m2, arch2, img) if m2 else None
-
-#         P1 = F.softmax(L1, dim=1)[0].cpu().numpy()
-#         top1_idx = int(P1.argmax())
-#         top1 = CLASS_NAMES[top1_idx]
-#         conf1 = float(P1[top1_idx])
-
-#         row = {"filename": base, "Model 1": model1_name, "M1:top": top1, "M1:conf": conf1}
-
-#         if L2 is not None:
-#             P2 = F.softmax(L2, dim=1)[0].cpu().numpy()
-#             top2_idx = int(P2.argmax())
-#             top2 = CLASS_NAMES[top2_idx]
-#             conf2 = float(P2[top2_idx])
-
-#             L_ens = ensemble_logits([L1, L2])
-#             P_ens = F.softmax(L_ens, dim=1)[0].cpu().numpy()
-#             topE_idx = int(P_ens.argmax())
-#             topE = CLASS_NAMES[topE_idx]
-#             confE = float(P_ens[topE_idx])
-
-#             row.update({"Model 2": model2_name, "M2:top": top2, "M2:conf": conf2, "ENS:top": topE, "ENS:conf": confE})
-
-#             if not compact:
-#                 for i, cls in enumerate(CLASS_NAMES):
-#                     row[f"M1:{cls}"]  = float(P1[i])
-#                     row[f"M2:{cls}"]  = float(P2[i])
-#                     row[f"ENS:{cls}"] = float(P_ens[i])
-#         else:
-#             if not compact:
-#                 for i, cls in enumerate(CLASS_NAMES):
-#                     row[f"M1:{cls}"] = float(P1[i])
-
-#         rows.append(row)
-
-#     return _round_df3(pd.DataFrame(rows))
-
 def predict_files(
     files: List[str],
     model1_name: str,
@@ -442,6 +375,22 @@ def _round_selected_cols(df: pd.DataFrame, cols: List[str]) -> pd.DataFrame:
             df[c] = pd.to_numeric(df[c], errors="coerce").round(3)
     return df
 
+# def load_best_singles_table(leader_csv: Path) -> pd.DataFrame:
+#     if not leader_csv.exists():
+#         return pd.DataFrame([{"info": "leaderboard.csv not found"}])
+#     df = pd.read_csv(leader_csv)
+#     if "source" in df.columns:
+#         df = df[df["source"] == "single_arch"].copy()
+#     sort_cols = [c for c in ["test_acc", "best_val_acc"] if c in df.columns]
+#     if sort_cols:
+#         df = df.sort_values(sort_cols, ascending=False)
+#     if "arch" in df.columns:
+#         df = df.drop_duplicates(subset=["arch"], keep="first")
+#     # Exclude run_dir from display
+#     keep = [c for c in ["arch","test_acc","best_val_acc","img_size"] if c in df.columns]
+#     df = df[keep].reset_index(drop=True) if keep else df
+#     return _round_selected_cols(df, ["test_acc","best_val_acc"])
+
 def load_best_singles_table(leader_csv: Path) -> pd.DataFrame:
     if not leader_csv.exists():
         return pd.DataFrame([{"info": "leaderboard.csv not found"}])
@@ -456,21 +405,43 @@ def load_best_singles_table(leader_csv: Path) -> pd.DataFrame:
     # Exclude run_dir from display
     keep = [c for c in ["arch","test_acc","best_val_acc","img_size"] if c in df.columns]
     df = df[keep].reset_index(drop=True) if keep else df
-    return _round_selected_cols(df, ["test_acc","best_val_acc"])
+
+    # ✅ CHANGE: Convert accuracy columns to percentages
+    if "test_acc" in df.columns:
+        df["test_acc"] = (df["test_acc"] * 100).round(2).astype(str) + "%"
+    if "best_val_acc" in df.columns:
+        df["best_val_acc"] = (df["best_val_acc"] * 100).round(2).astype(str) + "%"
+
+    return df
+
+# def load_best_pairs_table() -> Optional[pd.DataFrame]:
+#     if PAIR_COMBINED.exists():
+#         df = pd.read_csv(PAIR_COMBINED)
+#         show = [c for c in ["members_names","avg_acc","min_acc","acc_LukeTest","acc_MarcoTest","acc_FalahTest"] if c in df.columns]
+#         df = df[show].head(20) if show else df.head(20)
+#         return _round_selected_cols(df, ["avg_acc","min_acc","acc_LukeTest","acc_MarcoTest","acc_FalahTest"])
+#     #if PAIR_FALLBACK.exists():
+#     #    df = pd.read_csv(PAIR_FALLBACK)
+#     #    show = [c for c in ["members_names","acc"] if c in df.columns]
+#     #    df = df[show].head(20) if show else df.head(20)
+#     #    return _round_selected_cols(df, ["acc"])
+#     return None  # signal "not found"
 
 def load_best_pairs_table() -> Optional[pd.DataFrame]:
     if PAIR_COMBINED.exists():
         df = pd.read_csv(PAIR_COMBINED)
         show = [c for c in ["members_names","avg_acc","min_acc","acc_LukeTest","acc_MarcoTest","acc_FalahTest"] if c in df.columns]
         df = df[show].head(20) if show else df.head(20)
-        return _round_selected_cols(df, ["avg_acc","min_acc","acc_LukeTest","acc_MarcoTest","acc_FalahTest"])
-    #if PAIR_FALLBACK.exists():
-    #    df = pd.read_csv(PAIR_FALLBACK)
-    #    show = [c for c in ["members_names","acc"] if c in df.columns]
-    #    df = df[show].head(20) if show else df.head(20)
-    #    return _round_selected_cols(df, ["acc"])
-    return None  # signal "not found"
+        
+        # ✅ CHANGE: Convert accuracy columns to percentages
+        for col in ["avg_acc", "min_acc", "acc_LukeTest", "acc_MarcoTest", "acc_FalahTest"]:
+            if col in df.columns:
+                df[col] = (pd.to_numeric(df[col], errors="coerce") * 100).round(2).astype(str) + "%"
 
+        return df
+    return None
+
+    
 # ==============================
 # Upload helpers (gallery + state)
 # ==============================