removed hf-asr-leaderboard filter, improved parser, added query param dataset select option

app.py

@@ -14,7 +14,7 @@ def make_clickable(model_name):
 def get_model_ids():
     api = HfApi()
     # TODO: switch to hf-leaderboards for the final version.
-    models = api.list_models(filter="hf-asr-leaderboard")
+    models = api.list_models(filter="model-index")
     model_ids = [x.modelId for x in models]
     return model_ids
 
@@ -23,8 +23,8 @@ def get_metadata(model_id):
     try:
         readme_path = hf_hub_download(model_id, filename="README.md")
         return metadata_load(readme_path)
-    except requests.exceptions.HTTPError:
-        # 404 README.md not found
+    except Exception:
+        # 404 README.md not found or problem loading it
         return None
 
 
@@ -40,15 +40,15 @@ def parse_metric_value(value):
             value = value[0]
         else:
             value = None
-    value = round(value, 2) if value is not None else None
+    value = round(value, 2) if isinstance(value, float) else None
     return value
 
 
 def parse_metrics_rows(meta):
-    if "model-index" not in meta:
+    if not isinstance(meta["model-index"], list) or len(meta["model-index"]) == 0 or "results" not in meta["model-index"][0]:
         return None
     for result in meta["model-index"][0]["results"]:
-        if "dataset" not in result or "metrics" not in result:
+        if "dataset" not in result or "metrics" not in result or "type" not in result["dataset"]:
             continue
         dataset = result["dataset"]["type"]
         if "args" not in result["dataset"]:
@@ -56,7 +56,7 @@ def parse_metrics_rows(meta):
         row = {"dataset": dataset}
         for metric in result["metrics"]:
             type = metric["type"].lower().strip()
-            value = parse_metric_value(metric["value"])
+            value = parse_metric_value(metric.get("value", None))
             if value is None:
                 continue
             if type not in row or value < row[type]:
@@ -65,7 +65,7 @@ def parse_metrics_rows(meta):
         yield row
 
 
-@st.cache(ttl=600)
+@st.cache(ttl=86400)
 def get_data():
     data = []
     model_ids = get_model_ids()
@@ -86,10 +86,16 @@ selectable_datasets = list(set(dataframe.dataset.tolist()))
 
 st.markdown("# 🤗 Leaderboards")
 
+query_params = st.experimental_get_query_params()
+default_dataset = "common_voice"
+if "dataset" in query_params:
+    if len(query_params["dataset"]) > 0 and query_params["dataset"][0] in selectable_datasets:
+        default_dataset = query_params["dataset"][0]
+
 dataset = st.sidebar.selectbox(
     "Dataset",
     selectable_datasets,
-    index=selectable_datasets.index("common_voice"),
+    index=selectable_datasets.index(default_dataset),
 )
 
 dataset_df = dataframe[dataframe.dataset == dataset]