add sorting

app.py

@@ -19,6 +19,11 @@ search_backend = st.sidebar.selectbox(
     format_func=lambda x: {"hfapi": "Keyword search", "bm25": "BM25 search", "semantic": "Semantic Search"}[x],
 )
 limit_results = int(st.sidebar.number_input("Limit results", min_value=0, value=10))
+sort_by = st.sidebar.selectbox(
+    "Sort by",
+    [None, "downloads", "likes", "lastModified"],
+    format_func=lambda x: {None: "Relevance", "downloads": "Most downloads", "likes": "Most likes", "lastModified": "Recently updated"}[x],
+)
 
 st.sidebar.markdown("# Filters")
 args = ModelSearchArguments()
@@ -44,11 +49,11 @@ if search_query != "":
         "task": task,
     }
     if search_backend == "hfapi":
-        res = hf_api(search_query, limit_results, filters)
+        res = hf_api(search_query, limit_results, sort_by, filters)
     elif search_backend == "semantic":
-        res = semantic_search(search_query, limit_results, filters)
+        res = semantic_search(search_query, limit_results, sort_by, filters)
     elif search_backend == "bm25":
-        res = bm25_search(search_query, limit_results, filters)
+        res = bm25_search(search_query, limit_results, sort_by, filters)
     hit_list, hits_count = res["hits"], res["count"]
     hit_list = [
         {
@@ -79,7 +84,7 @@ if search_query != "":
             col3.metric("N° likes", numerize(hit["likes"]) if not math.isnan(hit["likes"]) else "N/A")
             st.button(
                 f"View model on 🤗",
-                on_click=lambda hit=hit: webbrowser.open(f"https://huggingface.co/{hit['modelId']}"),
+                on_click=lambda hit=hit: webbrowser.open(f"https://huggingface.co/{hit['modelId']}", new=2),
                 key=f"{i}-{hit['modelId']}",
             )
             st.write(f"**Tags:** {'  •  '.join(hit['tags'])}")

requirements.txt

@@ -2,4 +2,5 @@ pandas
 streamlit
 huggingface_hub
 numerize
-git+https://github.com/NouamaneTazi/hf_search@0.2
+pbr
+git+https://github.com/NouamaneTazi/hf_search@0.4

st_utils.py

@@ -5,20 +5,24 @@ from hf_search import HFSearch
 import streamlit as st
 import itertools
 
-hf_search = HFSearch(top_k=32)
+from pbr.version import VersionInfo
+print("hf_search version:", VersionInfo('hf_search').version_string())
+
+hf_search = HFSearch(top_k=1000)
 
 @st.cache
-def hf_api(query, limit=5, filters={}):
+def hf_api(query, limit=5, sort=None, filters={}):
     print("query", query)
     print("filters", filters)
     print("limit", limit)
+    print("sort", sort)
 
     api = HfApi()
     filt = ModelFilter(
         task=filters["task"],
         library=filters["library"],
     )
-    models = api.list_models(search=query, filter=filt, limit=limit, full=True)
+    models = api.list_models(search=query, filter=filt, limit=limit, sort=sort, full=True)
     hits = []
     for model in models:
         model = model.__dict__
@@ -37,12 +41,13 @@ def hf_api(query, limit=5, filters={}):
 
 
 @st.cache
-def semantic_search(query, limit=5, filters={}):
+def semantic_search(query, limit=5, sort=None, filters={}):
     print("query", query)
     print("filters", filters)
     print("limit", limit)
+    print("sort", sort)
 
-    hits = hf_search.search(query=query, method="retrieve & rerank", limit=limit, filters=filters)
+    hits = hf_search.search(query=query, method="retrieve & rerank", limit=limit, sort=sort, filters=filters)
     hits = [
         {
             "modelId": hit["modelId"],
@@ -57,13 +62,14 @@ def semantic_search(query, limit=5, filters={}):
 
 
 @st.cache
-def bm25_search(query, limit=5, filters={}):
+def bm25_search(query, limit=5, sort=None, filters={}):
     print("query", query)
     print("filters", filters)
     print("limit", limit)
+    print("sort", sort)
 
     # TODO: filters
-    hits = hf_search.search(query=query, method="bm25", limit=limit, filters=filters)
+    hits = hf_search.search(query=query, method="bm25", limit=limit, sort=sort, filters=filters)
     hits = [
         {
             "modelId": hit["modelId"],