fixed bugs in selection of verified results

app.py

@@ -57,9 +57,10 @@ def parse_metrics_rows(meta, from_autoeval=False):
             row["split"] = result["dataset"]["split"]
         if "config" in result["dataset"]:
             row["config"] = result["dataset"]["config"]
+        no_results = True
         for metric in result["metrics"]:
             type = metric["type"].lower().strip()
-            if type not in ("dataset", "split", "config", "verified"):
+            if type in ("model_id", "dataset", "split", "config", "verified"):
                 # Metrics are not allowed to be named "dataset", "split", "config", or "verified".
                 continue
             value = parse_metric_value(metric.get("value", None))
@@ -75,9 +76,13 @@ def parse_metrics_rows(meta, from_autoeval=False):
                     # it is a verified metric. Unverified metrics are already included
                     # in the leaderboard from the unverified model card.
                     if "verified" in metric and metric["verified"]:
+                        no_results = False
                         row[type] = value
                 else:
+                    no_results = False
                     row[type] = value
+        if no_results:
+            continue
         yield row
 
 @st.cache(ttl=3600)
@@ -144,7 +149,7 @@ dataset_df = dataset_df.dropna(axis="columns", how="all")
 if only_verified_results:
     dataset_df = dataset_df[dataset_df["verified"]]
 
-selectable_metrics = list(filter(lambda column: column not in ("model_id", "dataset"), dataset_df.columns))
+selectable_metrics = list(filter(lambda column: column not in ("model_id", "dataset", "split", "config", "verified"), dataset_df.columns))
 
 dataset_df = dataset_df.filter(["model_id"] + selectable_metrics)
 dataset_df = dataset_df.dropna(thresh=2)  # Want at least two non-na values (one for model_id and one for a metric).