update app

app.py

@@ -9,13 +9,14 @@ sys.path.append(str(path_root))
 
 
 encoder_index_map = {
-    'uniCOIL': ('UniCoil', 'index-unicoil'),
-    'SPLADE++ Ensemble Distil': ('SpladePlusPlusEnsembleDistil', 'index-splade-pp-ed'),
-    'SPLADE++ Self Distil': ('SpladePlusPlusSelfDistil', 'index-splade-pp-sd')
+    'uniCOIL': ('UniCoil', 'castorini/unicoil-noexp-msmarco-passage', 'index-unicoil'),
+    'SPLADE++ Ensemble Distil': ('SpladePlusPlusEnsembleDistil', 'naver/splade-cocondenser-ensembledistil', 'index-splade-pp-ed'),
+    'SPLADE++ Self Distil': ('SpladePlusPlusSelfDistil', 'naver/splade-cocondenser-ensembledistil', 'index-splade-pp-sd')
 }
 
 index = 'index-splade-pp-ed'
 encoder = 'SpladePlusPlusEnsembleDistil'
+encoder_index = 0
 
 st.set_page_config(page_title="Pyserini with ONNX Runtime",
                    page_icon='🌸', layout="centered")
@@ -24,14 +25,33 @@ cola, colb, colc = st.columns([5, 4, 5])
 with colb:
     st.image("logo.jpeg")
 
+
+colaa, colbb, colcc = st.columns([1, 8, 1])
+with colbb:
+    runtime = st.select_slider(
+        'Select a runtime type',
+        options=['PyTorch', 'ONNX Runtime'])
+    st.write('Now using: ', runtime)
+
+
 colaa, colbb, colcc = st.columns([1, 8, 1])
 with colbb:
     encoder = st.select_slider(
-        'Select a query encoder with ONNX Runtime',
+        'Select a query encoder',
         options=['uniCOIL', 'SPLADE++ Ensemble Distil', 'SPLADE++ Self Distil'])
     st.write('Now Running Encoder: ', encoder)
 
-encoder, index = encoder_index_map[encoder]
+if runtime == 'PyTorch':
+    runtime = 'pytorch'
+    runtime_index = 1
+else:
+    runtime = 'onnx'
+    runtime_index = 0
+
+encoder, index = encoder_index_map[encoder][runtime_index], encoder_index_map[encoder][2]
+
+searcher = LuceneImpactSearcher(
+    f'indexes/{index}', f'{encoder}', encoder_type=f'{runtime}')
 
 col1, col2 = st.columns([9, 1])
 with col1:
@@ -41,8 +61,6 @@ with col2:
     st.write('#')
     button_clicked = st.button("🔎")
 
-searcher = LuceneImpactSearcher(
-    f'indexes/{index}', f'{encoder}', encoder_type='onnx')
 
 if search_query or button_clicked:
     num_results = None
@@ -55,10 +73,13 @@ if search_query or button_clicked:
     for i, result in enumerate(search_results[:10]):
         result_score = result.score
         result_id = result.docid
-        output = f'<div class="row"> <div class="column"> <b>Rank</b>: {i+1} </div><div class="column"><b>Document ID</b>: {result_id}</div><div class="column"><b>Score</b>:{result_score:.2f}</div></div>'
+        contents = json.loads(result.raw)["contents"]
+        output = f'<div class="row"> <b>Rank</b>: {i+1} | <b>Document ID</b>: {result_id} | <b>Score</b>:{result_score:.2f}</div>'
 
         try:
             st.write(output, unsafe_allow_html=True)
+            st.write(
+                f'<div class="row">{contents}</div>', unsafe_allow_html=True)
 
         except:
             pass

pyserini/search/lucene/_impact_searcher.py

@@ -142,7 +142,7 @@ class LuceneImpactSearcher:
 
         encoded_query = self.encode(q)
 
-        jquery = encoded_query
+        jquery = JHashMap()
         if self.encoder_type == 'pytorch':
             for (token, weight) in encoded_query.items():
                 if token in self.idf and self.idf[token] > self.min_idf: