aikacl commited on
Commit
a231b39
Β·
verified Β·
1 Parent(s): 2b2c2cb

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +127 -79
app.py CHANGED
@@ -10,18 +10,80 @@ MODEL_ID = "KaLM-Embedding/KaLM-Reranker-V1-Nano"
10
  DEFAULT_INSTRUCTION = "Given a query, retrieve documents that answer the query."
11
  MAX_DOCS = 20
12
  MAX_DOC_CHARS = 4000
13
- DOC_SEPARATOR = "---"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
 
15
 
16
- def parse_documents(text: str) -> List[str]:
17
- docs = [doc.strip() for doc in text.split(DOC_SEPARATOR) if doc.strip()]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  docs = docs[:MAX_DOCS]
19
  docs = [doc[:MAX_DOC_CHARS] for doc in docs]
20
- return docs
21
 
 
 
 
 
22
 
23
- @spaces.GPU
24
- def rerank(query: str, documents: str, instruction: str, chunk_size: int):
25
  reranker = KaLMReranker(
26
  MODEL_ID,
27
  device=None,
@@ -34,12 +96,6 @@ def rerank(query: str, documents: str, instruction: str, chunk_size: int):
34
 
35
  query = query.strip()
36
  instruction = instruction.strip() or DEFAULT_INSTRUCTION
37
- docs = parse_documents(documents)
38
-
39
- if not query:
40
- return [], "Please input a query."
41
- if not docs:
42
- return [], "Please input at least one candidate document."
43
 
44
  try:
45
  rankings = reranker.rank(
@@ -64,49 +120,67 @@ def rerank(query: str, documents: str, instruction: str, chunk_size: int):
64
  return [], f"Error during reranking: {repr(error)}"
65
 
66
 
 
 
67
  with gr.Blocks(title="KaLM-Reranker-V1 Demo") as demo:
68
  gr.Markdown(
69
- f"""
70
  # KaLM-Reranker-V1 Demo
71
  **KaLM-Reranker-V1** is a fast but not late-interaction reranker for compressed document reranking.
72
  Input a query and several candidate documents. The demo returns relevance scores and reranked results.
73
- **Document format:** separate candidate documents with `{DOC_SEPARATOR}` (three dashes on its own line).
74
  """
75
  )
76
 
77
  with gr.Row():
 
78
  with gr.Column(scale=1):
79
- query = gr.Textbox(
80
  label="Query",
81
  value="What is the capital of China?",
82
  lines=3,
83
  )
84
- instruction = gr.Textbox(
85
  label="Instruction",
86
  value=DEFAULT_INSTRUCTION,
87
  lines=2,
88
  )
89
- chunk_size = gr.Dropdown(
90
  label="Chunk Size",
91
  choices=[2, 4, 8, 16],
92
  value=4,
93
  interactive=True,
94
  )
95
- documents = gr.Textbox(
96
- label="Candidate Documents",
97
- value=(
98
- "The capital of China is Beijing.\n"
99
- "---\n"
100
- "Gravity attracts bodies toward one another.\n"
101
- "---\n"
102
- "Shanghai is a major city in China.\n"
103
- "---\n"
104
- "Paris is the capital of France."
105
- ),
106
- lines=14,
107
  )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
108
  submit = gr.Button("Rerank", variant="primary")
109
 
 
110
  with gr.Column(scale=1):
111
  output_table = gr.Dataframe(
112
  headers=["Rank", "Corpus ID", "Score", "Document"],
@@ -115,74 +189,48 @@ Input a query and several candidate documents. The demo returns relevance scores
115
  )
116
  output_summary = gr.Markdown()
117
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
118
  submit.click(
119
  fn=rerank,
120
- inputs=[query, documents, instruction, chunk_size],
121
  outputs=[output_table, output_summary],
122
  )
123
 
124
- gr.Examples(
125
- examples=[
126
- [
127
- "What is the capital of China?",
128
- (
129
- "The capital of China is Beijing.\n"
130
- "---\n"
131
- "Gravity attracts bodies toward one another.\n"
132
- "---\n"
133
- "Paris is the capital of France."
134
- ),
135
- DEFAULT_INSTRUCTION,
136
- 4,
137
- ],
138
- [
139
- "Which model is suitable for efficient reranking?",
140
- (
141
- "KaLM-Reranker-V1-Nano is designed for efficient reranking.\n"
142
- "---\n"
143
- "Large language models are often expensive for reranking.\n"
144
- "---\n"
145
- "Image classifiers are used for visual recognition."
146
- ),
147
- DEFAULT_INSTRUCTION,
148
- 4,
149
- ],
150
- [
151
- "What is KaLM-Reranker-V1 designed for?",
152
- (
153
- "KaLM-Reranker-V1 is a reranker for compressed document reranking.\n"
154
- "---\n"
155
- "KaLM-Embedding is a general-purpose embedding model.\n"
156
- "---\n"
157
- "Weather forecasting predicts future weather conditions."
158
- ),
159
- DEFAULT_INSTRUCTION,
160
- 4,
161
- ],
162
- ],
163
- inputs=[query, documents, instruction, chunk_size],
164
- )
165
-
166
  gr.Markdown(
167
  """
168
  ## Citation
169
-
170
  If you find this demo useful, please cite:
171
-
172
  ```bibtex
173
  @misc{zhao2026kalmrerankerv1,
174
- title={KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking},
175
  author={Xinping Zhao and Jiaxin Xu and Ziqi Dai and Xin Zhang and Shouzheng Huang and Danyu Tang and Xinshuo Hu and Meishan Zhang and Baotian Hu and Min Zhang},
176
  year={2026},
177
  eprint={2606.22807},
178
  archivePrefix={arXiv},
179
  primaryClass={cs.CL},
180
- url={https://arxiv.org/abs/2606.22807},
181
- }
182
  ```
183
  """
184
  )
185
-
186
-
187
  if __name__ == "__main__":
188
  demo.launch()
 
10
  DEFAULT_INSTRUCTION = "Given a query, retrieve documents that answer the query."
11
  MAX_DOCS = 20
12
  MAX_DOC_CHARS = 4000
13
+ INIT_DOCS = 4
14
+
15
+ EXAMPLES = [
16
+ {
17
+ "label": "Capital of China",
18
+ "query": "What is the capital of China?",
19
+ "docs": [
20
+ "The capital of China is Beijing.",
21
+ "Gravity attracts bodies toward one another.",
22
+ "Paris is the capital of France.",
23
+ ],
24
+ },
25
+ {
26
+ "label": "Efficient Reranking",
27
+ "query": "Which model is suitable for efficient reranking?",
28
+ "docs": [
29
+ "KaLM-Reranker-V1-Nano is designed for efficient reranking.",
30
+ "Large language models are often expensive for reranking.",
31
+ "Image classifiers are used for visual recognition.",
32
+ ],
33
+ },
34
+ {
35
+ "label": "KaLM-Reranker Design",
36
+ "query": "What is KaLM-Reranker-V1 designed for?",
37
+ "docs": [
38
+ "KaLM-Reranker-V1 is a reranker for compressed document reranking.",
39
+ "KaLM-Embedding is a general-purpose embedding model.",
40
+ "Weather forecasting predicts future weather conditions.",
41
+ ],
42
+ },
43
+ ]
44
+
45
+
46
+ def _make_vis_updates(count: int):
47
+ return [gr.update(visible=(i < count)) for i in range(MAX_DOCS)]
48
+
49
+
50
+ def add_doc(count: int):
51
+ count = min(count + 1, MAX_DOCS)
52
+ return _make_vis_updates(count) + [count]
53
+
54
+
55
+ def remove_doc(count: int):
56
+ count = max(count - 1, 1)
57
+ return _make_vis_updates(count) + [count]
58
 
59
 
60
+ def load_example(example_idx):
61
+ if example_idx is None:
62
+ query_upd = gr.update(value="")
63
+ doc_upds = [gr.update(value="") for _ in range(MAX_DOCS)]
64
+ return [query_upd] + doc_upds + [INIT_DOCS]
65
+ ex = EXAMPLES[example_idx]
66
+ docs = ex["docs"]
67
+ count = len(docs)
68
+ query_upd = gr.update(value=ex["query"])
69
+ doc_upds = [
70
+ gr.update(value=docs[i] if i < count else "", visible=(i < count))
71
+ for i in range(MAX_DOCS)
72
+ ]
73
+ return [query_upd] + doc_upds + [count]
74
+
75
+
76
+ @spaces.GPU
77
+ def rerank(query: str, instruction: str, chunk_size: int, *doc_texts: str):
78
+ docs = [doc.strip() for doc in doc_texts if doc and doc.strip()]
79
  docs = docs[:MAX_DOCS]
80
  docs = [doc[:MAX_DOC_CHARS] for doc in docs]
 
81
 
82
+ if not query.strip():
83
+ return [], "Please input a query."
84
+ if not docs:
85
+ return [], "Please input at least one candidate document."
86
 
 
 
87
  reranker = KaLMReranker(
88
  MODEL_ID,
89
  device=None,
 
96
 
97
  query = query.strip()
98
  instruction = instruction.strip() or DEFAULT_INSTRUCTION
 
 
 
 
 
 
99
 
100
  try:
101
  rankings = reranker.rank(
 
120
  return [], f"Error during reranking: {repr(error)}"
121
 
122
 
123
+ # ── UI ──────────────────────────────────────────────────────────────────────
124
+
125
  with gr.Blocks(title="KaLM-Reranker-V1 Demo") as demo:
126
  gr.Markdown(
127
+ """
128
  # KaLM-Reranker-V1 Demo
129
  **KaLM-Reranker-V1** is a fast but not late-interaction reranker for compressed document reranking.
130
  Input a query and several candidate documents. The demo returns relevance scores and reranked results.
 
131
  """
132
  )
133
 
134
  with gr.Row():
135
+ # ── Left column: inputs ──
136
  with gr.Column(scale=1):
137
+ query_input = gr.Textbox(
138
  label="Query",
139
  value="What is the capital of China?",
140
  lines=3,
141
  )
142
+ instruction_input = gr.Textbox(
143
  label="Instruction",
144
  value=DEFAULT_INSTRUCTION,
145
  lines=2,
146
  )
147
+ chunk_size_input = gr.Dropdown(
148
  label="Chunk Size",
149
  choices=[2, 4, 8, 16],
150
  value=4,
151
  interactive=True,
152
  )
153
+
154
+ # Example selector
155
+ example_selector = gr.Dropdown(
156
+ label="Load Example",
157
+ choices=[ex["label"] for ex in EXAMPLES],
158
+ value=None,
159
+ interactive=True,
 
 
 
 
 
160
  )
161
+
162
+ gr.Markdown("### Candidate Documents")
163
+ doc_count = gr.State(value=INIT_DOCS)
164
+ doc_textboxes: List[gr.Textbox] = []
165
+
166
+ with gr.Column() as doc_list:
167
+ for i in range(MAX_DOCS):
168
+ visible = i < INIT_DOCS
169
+ tb = gr.Textbox(
170
+ label=f"Document {i + 1}",
171
+ lines=2,
172
+ placeholder=f"Enter document {i + 1}...",
173
+ visible=visible,
174
+ )
175
+ doc_textboxes.append(tb)
176
+
177
+ with gr.Row():
178
+ add_btn = gr.Button("+ Add Document", size="sm")
179
+ remove_btn = gr.Button("- Remove Last", size="sm")
180
+
181
  submit = gr.Button("Rerank", variant="primary")
182
 
183
+ # ── Right column: outputs ──
184
  with gr.Column(scale=1):
185
  output_table = gr.Dataframe(
186
  headers=["Rank", "Corpus ID", "Score", "Document"],
 
189
  )
190
  output_summary = gr.Markdown()
191
 
192
+ # ── Events ──
193
+ add_btn.click(
194
+ fn=add_doc,
195
+ inputs=[doc_count],
196
+ outputs=doc_textboxes + [doc_count],
197
+ )
198
+ remove_btn.click(
199
+ fn=remove_doc,
200
+ inputs=[doc_count],
201
+ outputs=doc_textboxes + [doc_count],
202
+ )
203
+ example_selector.change(
204
+ fn=load_example,
205
+ inputs=[example_selector],
206
+ outputs=[query_input] + doc_textboxes + [doc_count],
207
+ )
208
  submit.click(
209
  fn=rerank,
210
+ inputs=[query_input, instruction_input, chunk_size_input] + doc_textboxes,
211
  outputs=[output_table, output_summary],
212
  )
213
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
214
  gr.Markdown(
215
  """
216
  ## Citation
217
+
218
  If you find this demo useful, please cite:
219
+
220
  ```bibtex
221
  @misc{zhao2026kalmrerankerv1,
222
+ title={KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking},
223
  author={Xinping Zhao and Jiaxin Xu and Ziqi Dai and Xin Zhang and Shouzheng Huang and Danyu Tang and Xinshuo Hu and Meishan Zhang and Baotian Hu and Min Zhang},
224
  year={2026},
225
  eprint={2606.22807},
226
  archivePrefix={arXiv},
227
  primaryClass={cs.CL},
228
+ url={https://arxiv.org/abs/2606.22807},
229
+ }
230
  ```
231
  """
232
  )
233
+
234
+
235
  if __name__ == "__main__":
236
  demo.launch()