added task filtering

app.py

@@ -49,8 +49,8 @@ def parse_metrics_rows(meta, only_verified=False):
         for metric in result["metrics"]:
             name = metric["type"].lower().strip()
 
-            if name in ("model_id", "dataset", "split", "config"):
-                # Metrics are not allowed to be named "dataset", "split", "config".
+            if name in ("model_id", "dataset", "split", "config", "pipeline_tag", "only_verified"):
+                # Metrics are not allowed to be named "dataset", "split", "config", "pipeline_tag"
                 continue
             value = parse_metric_value(metric.get("value", None))
             if value is None:
@@ -88,36 +88,36 @@ def get_data_wrapper():
                 if row is None:
                     continue
                 row["model_id"] = model_info.id
+                row["pipeline_tag"] = model_info.pipeline_tag
+                row["only_verified"] = False
                 data.append(row)
             for row in parse_metrics_rows(meta, only_verified=True):
                 if row is None:
                     continue
                 row["model_id"] = model_info.id
-                verified_data.append(row)
+                row["pipeline_tag"] = model_info.pipeline_tag
+                row["only_verified"] = True
+                data.append(row)
         dataframe = pd.DataFrame.from_records(data)
         dataframe.to_pickle("cache.pkl")
-        verified_dataframe = pd.DataFrame.from_records(verified_data)
-        verified_dataframe.to_pickle("verified_cache.pkl")
 
-    if exists("cache.pkl") and exists("verified_cache.pkl"):
+    if exists("cache.pkl"):
         # If we have saved the results previously, call an asynchronous process
         # to fetch the results and update the saved file. Don't make users wait
         # while we fetch the new results. Instead, display the old results for
         # now. The new results should be loaded when this method
         # is called again.
         dataframe = pd.read_pickle("cache.pkl")
-        verified_dataframe = pd.read_pickle("verified_cache.pkl")
-        t = threading.Thread(name='get_data procs', target=get_data)
+        t = threading.Thread(name="get_data procs", target=get_data)
         t.start()
     else:
         # We have to make the users wait during the first startup of this app.
         get_data()
         dataframe = pd.read_pickle("cache.pkl")
-        verified_dataframe = pd.read_pickle("verified_cache.pkl")
 
-    return dataframe, verified_dataframe
+    return dataframe
 
-dataframe, verified_dataframe = get_data_wrapper()
+dataframe = get_data_wrapper()
 
 st.markdown("# 🤗 Leaderboards")
 
@@ -125,17 +125,29 @@ only_verified_results = st.sidebar.checkbox(
     "Filter for Verified Results",
 )
 
-selectable_datasets = sorted(list(set(dataframe.dataset.tolist())), key=lambda name: name.lower())
+selectable_tasks = list(set(dataframe.pipeline_tag))
+if None in selectable_tasks:
+    selectable_tasks.remove(None)
+selectable_tasks.sort(key=lambda name: name.lower())
+
+task = st.sidebar.selectbox(
+    "Task",
+    ["-any-"] + selectable_tasks,
+)
 
-if only_verified_results:
-    dataframe = verified_dataframe
+if task != "-any-":
+    dataframe = dataframe[dataframe.pipeline_tag == task]
+
+selectable_datasets = sorted(list(set(dataframe.dataset.tolist())), key=lambda name: name.lower())
+if "" in selectable_datasets:
+    selectable_datasets.remove("")
 
 query_params = st.experimental_get_query_params()
 if "first_query_params" not in st.session_state:
     st.session_state.first_query_params = query_params
 first_query_params = st.session_state.first_query_params
 
-default_dataset = "common_voice"
+default_dataset = None
 if "dataset" in first_query_params:
     if len(first_query_params["dataset"]) > 0 and first_query_params["dataset"][0] in selectable_datasets:
         default_dataset = first_query_params["dataset"][0]
@@ -143,9 +155,11 @@ if "dataset" in first_query_params:
 dataset = st.sidebar.selectbox(
     "Dataset",
     selectable_datasets,
-    index=selectable_datasets.index(default_dataset),
+    index=selectable_datasets.index(default_dataset) if default_dataset in selectable_datasets else 0,
 )
 
+dataframe = dataframe[dataframe.only_verified == only_verified_results]
+
 st.experimental_set_query_params(**{"dataset": [dataset]})
 
 dataset_df = dataframe[dataframe.dataset == dataset]
@@ -167,7 +181,7 @@ if len(dataset_df) > 0:
     )
     dataset_df = dataset_df[dataset_df.split == split]
 
-    selectable_metrics = list(filter(lambda column: column not in ("model_id", "dataset", "split", "config"), dataset_df.columns))
+    selectable_metrics = list(filter(lambda column: column not in ("model_id", "dataset", "split", "config", "pipeline_tag", "only_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).
@@ -217,5 +231,5 @@ if len(dataset_df) > 0:
 
 else:
     st.markdown(
-        "No data to display."
+        "No " + ("verified" if only_verified_results else "unverified") + " results to display. Try toggling the verified results filter."
     )