Spaces:
Runtime error
Runtime error
wip
Browse files
app.py
CHANGED
@@ -1,29 +1,59 @@
|
|
1 |
import gradio as gr
|
2 |
import ir_datasets
|
3 |
import pandas as pd
|
|
|
4 |
|
5 |
from autogluon.multimodal import MultiModalPredictor
|
6 |
|
7 |
|
8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
model_name = "sentence-transformers/all-MiniLM-L6-v2"
|
10 |
-
# dataset = ir_datasets.load("beir/fiqa/dev")
|
11 |
-
# docs_df = pd.DataFrame(dataset.docs_iter()).set_index("doc_id").sample(frac=0.001)
|
12 |
predictor = MultiModalPredictor(
|
13 |
pipeline="feature_extraction",
|
14 |
hyperparameters={
|
15 |
"model.hf_text.checkpoint_name": model_name
|
16 |
}
|
17 |
)
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
def main():
|
25 |
with gr.Blocks(title="OpenSearch Demo") as demo:
|
26 |
-
gr.Markdown("#
|
27 |
gr.Markdown("Ask an open question!")
|
28 |
with gr.Row():
|
29 |
inp_single = gr.Textbox(show_label=False)
|
@@ -31,23 +61,25 @@ def main():
|
|
31 |
btn_single = gr.Button("Generate Embedding")
|
32 |
with gr.Row():
|
33 |
out_single = gr.DataFrame(label="Embedding", show_label=True)
|
34 |
-
gr.Markdown("You can select one of the sample datasets for
|
35 |
-
with gr.Row():
|
36 |
-
with gr.Column():
|
37 |
-
btn_fiqa = gr.Button("fiqa")
|
38 |
-
with gr.Column():
|
39 |
-
btn_faiss = gr.Button("faiss")
|
40 |
with gr.Row():
|
41 |
-
|
42 |
-
gr.Markdown("You can also try out our batch inference by uploading a file")
|
43 |
with gr.Row():
|
44 |
-
out_batch = gr.
|
|
|
45 |
with gr.Row():
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
demo.launch()
|
52 |
|
53 |
|
|
|
1 |
import gradio as gr
|
2 |
import ir_datasets
|
3 |
import pandas as pd
|
4 |
+
import numpy as np
|
5 |
|
6 |
from autogluon.multimodal import MultiModalPredictor
|
7 |
|
8 |
|
9 |
+
query_embedding = None
|
10 |
+
document_embedding = None
|
11 |
+
docs_df = None
|
12 |
+
|
13 |
+
def text_embedding_batch():
|
14 |
+
model_name = "sentence-transformers/all-MiniLM-L6-v2"
|
15 |
+
dataset = ir_datasets.load("beir/fiqa/dev")
|
16 |
+
docs_df = pd.DataFrame(dataset.docs_iter()).set_index("doc_id").sample(frac=0.0001)
|
17 |
+
predictor = MultiModalPredictor(
|
18 |
+
pipeline="feature_extraction",
|
19 |
+
hyperparameters={
|
20 |
+
"model.hf_text.checkpoint_name": model_name
|
21 |
+
}
|
22 |
+
)
|
23 |
+
embedding = predictor.extract_embedding(docs_df)
|
24 |
+
query_embedding = embedding["text"]
|
25 |
+
return query_embedding
|
26 |
+
|
27 |
+
|
28 |
+
def text_embedding_single(query: str):
|
29 |
model_name = "sentence-transformers/all-MiniLM-L6-v2"
|
|
|
|
|
30 |
predictor = MultiModalPredictor(
|
31 |
pipeline="feature_extraction",
|
32 |
hyperparameters={
|
33 |
"model.hf_text.checkpoint_name": model_name
|
34 |
}
|
35 |
)
|
36 |
+
embedding = predictor.extract_embedding([query])
|
37 |
+
document_embedding = embedding["0"]
|
38 |
+
return document_embedding
|
39 |
+
|
40 |
+
|
41 |
+
def rank_document():
|
42 |
+
q_norm = query_embedding / np.linalg.norm(query_embedding, axis=-1, keepdims=True)
|
43 |
+
print(q_norm)
|
44 |
+
d_norm = document_embedding / np.linalg.norm(document_embedding, axis=-1, keepdims=True)
|
45 |
+
scores = d_norm.dot(q_norm[0])
|
46 |
+
print(scores)
|
47 |
|
48 |
+
result = []
|
49 |
+
for idx in np.argsort(-scores)[:2]:
|
50 |
+
result.append(docs_df['text'].iloc[idx])
|
51 |
+
return result
|
52 |
+
|
53 |
|
54 |
def main():
|
55 |
with gr.Blocks(title="OpenSearch Demo") as demo:
|
56 |
+
gr.Markdown("# Semantic Search with Autogluon")
|
57 |
gr.Markdown("Ask an open question!")
|
58 |
with gr.Row():
|
59 |
inp_single = gr.Textbox(show_label=False)
|
|
|
61 |
btn_single = gr.Button("Generate Embedding")
|
62 |
with gr.Row():
|
63 |
out_single = gr.DataFrame(label="Embedding", show_label=True)
|
64 |
+
gr.Markdown("You can select one of the sample datasets for document embedding")
|
|
|
|
|
|
|
|
|
|
|
65 |
with gr.Row():
|
66 |
+
btn_fiqa = gr.Button("fiqa")
|
|
|
67 |
with gr.Row():
|
68 |
+
out_batch = gr.DataFrame(label="Sample Embeddings", show_label=True, row_count=5)
|
69 |
+
gr.Markdown("Now rank the documents and pick the top 3 most relevant from the dataset")
|
70 |
with gr.Row():
|
71 |
+
btn_rank = gr.Button("Rank documents")
|
72 |
+
with gr.Row():
|
73 |
+
out_rank = gr.DataFrame(label="Top ranked documents", show_label=True, row_count=5)
|
74 |
+
# with gr.Row():
|
75 |
+
# out_batch = gr.File(interactive=True)
|
76 |
+
# with gr.Row():
|
77 |
+
# btn_file = gr.Button("Generate Embedding")
|
78 |
+
|
79 |
+
btn_single.click(fn=text_embedding_single, inputs=inp_single, outputs=out_single)
|
80 |
+
btn_fiqa.click(fn=text_embedding_batch, inputs=None, outputs=out_batch)
|
81 |
+
btn_rank.click(fn=rank_document, inputs=None, outputs=out_rank)
|
82 |
+
# btn_file.click(fn=text_embedding_batch, inputs=inp_single, outputs=out_single)
|
83 |
demo.launch()
|
84 |
|
85 |
|