Refactor glob to use the cache of HfFileSystem

src/details.py

@@ -2,10 +2,9 @@ import asyncio
 
 import gradio as gr
 import pandas as pd
-from huggingface_hub import HfFileSystem
 
 import src.constants as constants
-from src.hub import load_jsonlines_file
+from src.hub import glob, load_jsonlines_file
 
 
 def update_task_description_component(task):
@@ -40,16 +39,18 @@ def update_load_details_component(model_id_1, model_id_2, subtask):
         return gr.Button("Load Details", interactive=False)
 
 
+def fetch_details_paths(model_id, subtask):
+    model_name_sanitized = model_id.replace("/", "__")
+    dataset_id = constants.DETAILS_DATASET_ID.format(model_name_sanitized=model_name_sanitized)
+    filename = constants.DETAILS_FILENAME.format(subtask=subtask)
+    path = f"{dataset_id}/**/{filename}"
+    return glob(path)
+
+
 async def load_details_dataframe(model_id, subtask):
-    fs = HfFileSystem()
     if not model_id or not subtask:
         return
-    model_name_sanitized = model_id.replace("/", "__")
-    paths = fs.glob(
-        f"{constants.DETAILS_DATASET_ID}/**/{constants.DETAILS_FILENAME}".format(
-            model_name_sanitized=model_name_sanitized, subtask=subtask
-        )
-    )
+    paths = fetch_details_paths(model_id, subtask)
     if not paths:
         return
     path = max(paths)

src/hub.py

@@ -2,11 +2,17 @@ import io
 import json
 
 import httpx
-from huggingface_hub import hf_hub_url
+from huggingface_hub import HfFileSystem, hf_hub_url
 from huggingface_hub.utils import build_hf_headers
 
 
 client = httpx.AsyncClient(follow_redirects=True)
+fs = HfFileSystem()
+
+
+def glob(path):
+    paths = fs.glob(path)
+    return paths
 
 
 async def load_json_file(path):

src/results.py

@@ -2,16 +2,14 @@ import asyncio
 
 import gradio as gr
 import pandas as pd
-from huggingface_hub import HfFileSystem
 
 import src.constants as constants
-from src.hub import load_json_file
+from src.hub import glob, load_json_file
 
 
 def fetch_result_paths():
-    fs = HfFileSystem()
-    paths = fs.glob(f"{constants.RESULTS_DATASET_ID}/**/**/*.json")
-    return paths
+    path = f"{constants.RESULTS_DATASET_ID}/**/**/*.json"
+    return glob(path)
 
 
 def sort_result_paths_per_model(paths):