updates

app.py

@@ -218,22 +218,14 @@ def change_tab(query_param: str):
 # Searching and filtering
 def update_table(hidden_df: pd.DataFrame, current_columns_df: pd.DataFrame, columns: list, type_query: list, size_query: list, show_deleted: bool, query: str):
     filtered_df = filter_models(hidden_df, type_query, size_query, show_deleted)
-    df = search_table(filtered_df, current_columns_df, query)
-    df = select_columns(df, columns)
+    if query != "":
+        filtered_df = search_table(filtered_df, query)
+    df = select_columns(filtered_df, columns)
 
     return df
 
-def search_table(df: pd.DataFrame, current_columns_df: pd.DataFrame, query: str) -> pd.DataFrame:
-    current_columns = current_columns_df.columns
-    if AutoEvalColumn.model_type.name in current_columns:
-        filtered_df = df[
-            (df[AutoEvalColumn.dummy.name].str.contains(query, case=False))
-            | (df[AutoEvalColumn.model_type.name].str.contains(query, case=False))
-        ]
-    else:
-        filtered_df = df[(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))]
-
-    return filtered_df
+def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
+    return df[(df[AutoEvalColumn.dummy.name].str.contains(query, case=False))]
 
 def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
     always_here_cols = [
@@ -247,12 +239,13 @@ def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
     return filtered_df
 
 NUMERIC_INTERVALS = {
-    "< 1.5B": (0, 1.5),
-    "~3B": (1.5, 5),
-    "~7B": (6, 11),
-    "~13B": (12, 15),
-    "~35B": (16, 55),
-    "60B+": (55, 10000),
+    "Unknown": pd.Interval(-1, 0, closed="right"), 
+    "< 1.5B": pd.Interval(0, 1.5, closed="right"),
+    "~3B": pd.Interval(1.5, 5, closed="right"),
+    "~7B": pd.Interval(6, 11, closed="right"),
+    "~13B": pd.Interval(12, 15, closed="right"),
+    "~35B": pd.Interval(16, 55, closed="right"),
+    "60B+": pd.Interval(55, 10000, closed="right"),
 }
 
 def filter_models(
@@ -267,9 +260,10 @@ def filter_models(
     type_emoji = [t[0] for t in type_query]
     filtered_df = filtered_df[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
 
-    numeric_interval = [NUMERIC_INTERVALS[s] for s in size_query]
+    numeric_interval = pd.IntervalIndex(sorted([NUMERIC_INTERVALS[s] for s in size_query]))
     params_column = pd.to_numeric(df[AutoEvalColumn.params.name], errors="coerce")
-    filtered_df = filtered_df.loc[params_column.between(numeric_interval[0][0], numeric_interval[-1][1])]
+    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
+    filtered_df = filtered_df.loc[mask]
 
     return filtered_df
 

model_info_cache.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c307938f15bda18b6c38af3d02cc0407d9d8d5345bc31f475af2cbbb33a4f8b5
+size 2895750

src/display_models/get_model_metadata.py

@@ -2,6 +2,7 @@ import glob
 import json
 import os
 import re
+import pickle
 from typing import List
 
 import huggingface_hub
@@ -16,27 +17,43 @@ api = HfApi(token=os.environ.get("H4_TOKEN", None))
 
 
 def get_model_infos_from_hub(leaderboard_data: List[dict]):
+    # load cache from disk
+    try:
+        with open("model_info_cache.pkl", "rb") as f:
+            model_info_cache = pickle.load(f)
+    except EOFError:
+        model_info_cache = {}
+
     for model_data in tqdm(leaderboard_data):
         model_name = model_data["model_name_for_query"]
-        try:
-            model_info = api.model_info(model_name)
-        except huggingface_hub.utils._errors.RepositoryNotFoundError:
-            print("Repo not found!", model_name)
-            model_data[AutoEvalColumn.license.name] = None
-            model_data[AutoEvalColumn.likes.name] = None
-            model_data[AutoEvalColumn.params.name] = get_model_size(model_name, None)
-            continue
+
+        if model_name in model_info_cache:
+            model_info = model_info_cache[model_name]
+        else:
+            try:
+                model_info = api.model_info(model_name)
+                model_info_cache[model_name] = model_info
+            except huggingface_hub.utils._errors.RepositoryNotFoundError:
+                print("Repo not found!", model_name)
+                model_data[AutoEvalColumn.license.name] = None
+                model_data[AutoEvalColumn.likes.name] = None
+                model_data[AutoEvalColumn.params.name] = get_model_size(model_name, None)
+                continue
 
         model_data[AutoEvalColumn.license.name] = get_model_license(model_info)
         model_data[AutoEvalColumn.likes.name] = get_model_likes(model_info)
         model_data[AutoEvalColumn.params.name] = get_model_size(model_name, model_info)
+    
+    # save cache to disk in pickle format
+    with open("model_info_cache.pkl", "wb") as f:
+        pickle.dump(model_info_cache, f)
 
 
 def get_model_license(model_info):
     try:
         return model_info.cardData["license"]
     except Exception:
-        return None
+        return "?"
 
 
 def get_model_likes(model_info):
@@ -56,7 +73,7 @@ def get_model_size(model_name, model_info):
             size = size_match.group(0)
             return round(float(size[:-1]) if size[-1] == "b" else float(size[:-1]) / 1e3, 3)
         except AttributeError:
-            return None
+            return 0
 
 
 def get_model_type(leaderboard_data: List[dict]):

src/display_models/model_metadata_type.py

@@ -22,6 +22,8 @@ class ModelType(Enum):
 
 MODEL_TYPE_METADATA: Dict[str, ModelType] = {
     "tiiuae/falcon-180B": ModelType.PT,
+    "tiiuae/falcon-180B-chat": ModelType.RL,
+    "microsoft/phi-1_5": ModelType.PT,
     "Qwen/Qwen-7B": ModelType.PT,
     "Qwen/Qwen-7B-Chat": ModelType.RL,
     "notstoic/PygmalionCoT-7b": ModelType.IFT,

src/display_models/read_results.py

@@ -27,7 +27,7 @@ class EvalResult:
     results: dict
     precision: str = ""
     model_type: str = ""
-    weight_type: str = ""
+    weight_type: str = "Original"
     date: str = ""
 
     def to_dict(self):