async version with more cachine

app.py

@@ -3,23 +3,27 @@ import copy
 import os
 from dataclasses import asdict, dataclass
 from datetime import datetime, timedelta
+from functools import lru_cache
 from json import JSONDecodeError
 from typing import Any, Dict, List, Optional, Union
+
 import gradio as gr
 import httpx
 import orjson
+from cachetools import TTLCache, cached
 from cashews import NOT_NONE, cache
-from httpx import AsyncClient
+from httpx import AsyncClient, Client
 from huggingface_hub import hf_hub_url, logging
 from huggingface_hub.utils import disable_progress_bars
 from rich import print
 from tqdm.auto import tqdm
-from httpx import Client
-from datetime import datetime, timedelta
 
-cache.setup(
-    "mem://"
-) 
+CACHE_EXPIRY_TIME = timedelta(hours=3)
+
+sync_cache = TTLCache(maxsize=200_000, ttl=CACHE_EXPIRY_TIME, timer=datetime.now)
+
+
+cache.setup("mem://")
 
 
 disable_progress_bars()
@@ -58,6 +62,7 @@ async def _try_load_model_card(hub_id, client=None):
         length = None
     return card_text, length
 
+
 def _try_parse_card_data(hub_json_data):
     data = {}
     keys = ["license", "language", "datasets"]
@@ -72,7 +77,7 @@ def _try_parse_card_data(hub_json_data):
     return data
 
 
-@dataclass
+@dataclass(eq=False)
 class ModelMetadata:
     hub_id: str
     tags: Optional[List[str]]
@@ -89,7 +94,7 @@ class ModelMetadata:
     created_at: Optional[datetime] = None
 
     @classmethod
-    @cache(ttl=timedelta(hours=3), condition=NOT_NONE)
+    @cache(ttl=CACHE_EXPIRY_TIME, condition=NOT_NONE)
     async def from_hub(cls, hub_id, client=None):
         try:
             if not client:
@@ -224,6 +229,7 @@ ALL_PIPELINES = {
 }
 
 
+@lru_cache()
 def generate_task_scores_dict():
     task_scores = {}
     for task in ALL_PIPELINES:
@@ -262,6 +268,7 @@ def generate_task_scores_dict():
     return task_scores
 
 
+@lru_cache()
 def generate_common_scores():
     GENERIC_SCORES = copy.deepcopy(COMMON_SCORES)
     GENERIC_SCORES["_max_score"] = sum(
@@ -274,6 +281,7 @@ SCORES = generate_task_scores_dict()
 GENERIC_SCORES = generate_common_scores()
 
 
+@cached(sync_cache)
 def _basic_check(data: Optional[ModelMetadata]):
     score = 0
     if data is None:
@@ -334,7 +342,7 @@ def create_query_url(query, skip=0):
     return f"https://huggingface.co/api/search/full-text?q={query}&limit=100&skip={skip}&type=model"
 
 
-def get_results(query,sync_client=None) -> Dict[Any, Any]:
+def get_results(query, sync_client=None) -> Dict[Any, Any]:
     if not sync_client:
         sync_client = Client(http2=True, headers=headers)
     url = create_query_url(query)
@@ -461,6 +469,7 @@ def create_markdown(results):  # TODO move to separate file
         rows.append(row)
     return "\n".join(rows)
 
+
 async def get_result_card_snippet(result, query=None, client=None):
     if not client:
         client = AsyncClient(http2=True, headers=headers)
@@ -472,6 +481,7 @@ async def get_result_card_snippet(result, query=None, client=None):
         result["text"] = "Could not load model card"
     return result
 
+
 @cache(ttl=timedelta(hours=3), condition=NOT_NONE)
 async def get_result_card_snippets(results, query=None, client=None):
     if not client:
@@ -483,8 +493,10 @@ async def get_result_card_snippets(results, query=None, client=None):
     results = await asyncio.gather(*result_snippets)
     return results
 
+
 sync_client = Client(http2=True, headers=headers)
 
+
 def _search_hub(
     query: str,
     min_score: Optional[int] = None,

requirements.txt

@@ -30,6 +30,8 @@ attrs==23.1.0
     #   jsonschema
 backcall==0.2.0
     # via ipython
+cachetools==5.3.1
+    # via -r requirements.in
 cashews==6.2.0
     # via -r requirements.in
 certifi==2023.5.7