Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -10,17 +10,18 @@ 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 |
|
| 14 |
|
| 15 |
def parse_documents(text: str) -> List[str]:
|
| 16 |
-
docs = [doc.strip() for doc in text.split(
|
| 17 |
docs = docs[:MAX_DOCS]
|
| 18 |
docs = [doc[:MAX_DOC_CHARS] for doc in docs]
|
| 19 |
return docs
|
| 20 |
|
| 21 |
|
| 22 |
@spaces.GPU
|
| 23 |
-
def rerank(query: str, documents: str, instruction: str):
|
| 24 |
reranker = KaLMReranker(
|
| 25 |
MODEL_ID,
|
| 26 |
device=None,
|
|
@@ -28,7 +29,7 @@ def rerank(query: str, documents: str, instruction: str):
|
|
| 28 |
batch_size=4,
|
| 29 |
query_max_length=512,
|
| 30 |
max_length=1024,
|
| 31 |
-
chunk_size=
|
| 32 |
)
|
| 33 |
|
| 34 |
query = query.strip()
|
|
@@ -56,7 +57,7 @@ def rerank(query: str, documents: str, instruction: str):
|
|
| 56 |
|
| 57 |
summary = (
|
| 58 |
f"Reranked {len(docs)} documents with "
|
| 59 |
-
f"`{MODEL_ID}`. Higher score means more relevant."
|
| 60 |
)
|
| 61 |
return table, summary
|
| 62 |
except Exception as error:
|
|
@@ -65,11 +66,11 @@ def rerank(query: str, documents: str, instruction: str):
|
|
| 65 |
|
| 66 |
with gr.Blocks(title="KaLM-Reranker-V1 Demo") as demo:
|
| 67 |
gr.Markdown(
|
| 68 |
-
"""
|
| 69 |
# KaLM-Reranker-V1 Demo
|
| 70 |
**KaLM-Reranker-V1** is a fast but not late-interaction reranker for compressed document reranking.
|
| 71 |
Input a query and several candidate documents. The demo returns relevance scores and reranked results.
|
| 72 |
-
**Document format:** separate candidate documents with
|
| 73 |
"""
|
| 74 |
)
|
| 75 |
|
|
@@ -85,12 +86,21 @@ Input a query and several candidate documents. The demo returns relevance scores
|
|
| 85 |
value=DEFAULT_INSTRUCTION,
|
| 86 |
lines=2,
|
| 87 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
documents = gr.Textbox(
|
| 89 |
label="Candidate Documents",
|
| 90 |
value=(
|
| 91 |
-
"The capital of China is Beijing.\n
|
| 92 |
-
"
|
| 93 |
-
"
|
|
|
|
|
|
|
|
|
|
| 94 |
"Paris is the capital of France."
|
| 95 |
),
|
| 96 |
lines=14,
|
|
@@ -107,7 +117,7 @@ Input a query and several candidate documents. The demo returns relevance scores
|
|
| 107 |
|
| 108 |
submit.click(
|
| 109 |
fn=rerank,
|
| 110 |
-
inputs=[query, documents, instruction],
|
| 111 |
outputs=[output_table, output_summary],
|
| 112 |
)
|
| 113 |
|
|
@@ -116,32 +126,41 @@ Input a query and several candidate documents. The demo returns relevance scores
|
|
| 116 |
[
|
| 117 |
"What is the capital of China?",
|
| 118 |
(
|
| 119 |
-
"The capital of China is Beijing.\n
|
| 120 |
-
"
|
|
|
|
|
|
|
| 121 |
"Paris is the capital of France."
|
| 122 |
),
|
| 123 |
DEFAULT_INSTRUCTION,
|
|
|
|
| 124 |
],
|
| 125 |
[
|
| 126 |
"Which model is suitable for efficient reranking?",
|
| 127 |
(
|
| 128 |
-
"KaLM-Reranker-V1-Nano is designed for efficient reranking.\n
|
| 129 |
-
"
|
|
|
|
|
|
|
| 130 |
"Image classifiers are used for visual recognition."
|
| 131 |
),
|
| 132 |
DEFAULT_INSTRUCTION,
|
|
|
|
| 133 |
],
|
| 134 |
[
|
| 135 |
"What is KaLM-Reranker-V1 designed for?",
|
| 136 |
(
|
| 137 |
-
"KaLM-Reranker-V1 is a reranker for compressed document reranking.\n
|
| 138 |
-
"
|
|
|
|
|
|
|
| 139 |
"Weather forecasting predicts future weather conditions."
|
| 140 |
),
|
| 141 |
DEFAULT_INSTRUCTION,
|
|
|
|
| 142 |
],
|
| 143 |
],
|
| 144 |
-
inputs=[query, documents, instruction],
|
| 145 |
)
|
| 146 |
|
| 147 |
gr.Markdown(
|
|
|
|
| 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,
|
|
|
|
| 29 |
batch_size=4,
|
| 30 |
query_max_length=512,
|
| 31 |
max_length=1024,
|
| 32 |
+
chunk_size=chunk_size,
|
| 33 |
)
|
| 34 |
|
| 35 |
query = query.strip()
|
|
|
|
| 57 |
|
| 58 |
summary = (
|
| 59 |
f"Reranked {len(docs)} documents with "
|
| 60 |
+
f"`{MODEL_ID}` (chunk_size={chunk_size}). Higher score means more relevant."
|
| 61 |
)
|
| 62 |
return table, summary
|
| 63 |
except Exception as error:
|
|
|
|
| 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 |
|
|
|
|
| 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,
|
|
|
|
| 117 |
|
| 118 |
submit.click(
|
| 119 |
fn=rerank,
|
| 120 |
+
inputs=[query, documents, instruction, chunk_size],
|
| 121 |
outputs=[output_table, output_summary],
|
| 122 |
)
|
| 123 |
|
|
|
|
| 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(
|