Spaces:
Runtime error
Runtime error
orionweller
commited on
Commit
•
7eba807
1
Parent(s):
61f7243
changes
Browse files- app.py +26 -35
- requirements.txt +3 -0
app.py
CHANGED
@@ -1,56 +1,40 @@
|
|
1 |
import gradio as gr
|
2 |
from sentence_transformers import SentenceTransformer, util, CrossEncoder
|
3 |
-
import torch
|
4 |
from transformers import set_seed
|
5 |
import numpy as np
|
6 |
-
import pandas as pd
|
7 |
-
import argparse
|
8 |
|
9 |
set_seed(42)
|
10 |
|
|
|
11 |
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
"""
|
14 |
Input:
|
15 |
doc1, doc2: strings containing the documents/passages
|
16 |
query1, query2: strings for queries that are only relevant to the corresponding doc (doc1 -> q1, doc2 -> q2)
|
17 |
model_name: string containing the type of model to run
|
18 |
-
model: the preloaded model, if caching
|
19 |
|
20 |
Returns:
|
21 |
A dictionary containing each query (q1 or q2) and the score (P@1) for the pair
|
22 |
|
23 |
"""
|
24 |
-
### Model initialization
|
25 |
-
if model_name == "dpr":
|
26 |
-
model_type = "dpr"
|
27 |
-
if model is not None:
|
28 |
-
passage_encoder, query_encoder = model
|
29 |
-
else:
|
30 |
-
passage_encoder = SentenceTransformer(
|
31 |
-
"facebook-dpr-ctx_encoder-multiset-base"
|
32 |
-
)
|
33 |
-
query_encoder = SentenceTransformer(
|
34 |
-
"facebook-dpr-question_encoder-multiset-base"
|
35 |
-
)
|
36 |
-
elif "cross-encoder" in model_name or "t5" in model_name:
|
37 |
-
model_type = "cross_encoder"
|
38 |
-
if model is None:
|
39 |
-
model = CrossEncoder(model_name)
|
40 |
-
else:
|
41 |
-
model_type = "biencoder"
|
42 |
-
if model is not None:
|
43 |
-
embedder = model
|
44 |
-
else:
|
45 |
-
embedder = SentenceTransformer(model_name)
|
46 |
-
|
47 |
corpus = [doc1, doc2]
|
48 |
queries = [q1, q2]
|
49 |
results = {}
|
50 |
num_correct = 0
|
51 |
|
52 |
### Do Retrieval
|
53 |
-
if
|
54 |
passage_embeddings = passage_encoder.encode(corpus)
|
55 |
|
56 |
query_encoder = SentenceTransformer(
|
@@ -69,7 +53,7 @@ def calc_preferred_dense(doc1, doc2, q1, q2, model_name="dpr", model=None):
|
|
69 |
num_correct += 1
|
70 |
model = (passage_encoder, query_encoder)
|
71 |
|
72 |
-
elif
|
73 |
for idx, query in enumerate(queries):
|
74 |
scores = model.predict([[query, doc1], [query, doc2]])
|
75 |
results[f"q{idx+1}"] = scores.tolist()
|
@@ -100,11 +84,18 @@ def calc_preferred_dense(doc1, doc2, q1, q2, model_name="dpr", model=None):
|
|
100 |
model = embedder
|
101 |
|
102 |
results["score"] = num_correct / 2
|
103 |
-
return results
|
104 |
|
105 |
|
106 |
-
|
107 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
|
109 |
-
iface = gr.Interface(fn=calc_preferred_dense, inputs="text", outputs="text")
|
110 |
-
iface.launch()
|
|
|
1 |
import gradio as gr
|
2 |
from sentence_transformers import SentenceTransformer, util, CrossEncoder
|
|
|
3 |
from transformers import set_seed
|
4 |
import numpy as np
|
|
|
|
|
5 |
|
6 |
set_seed(42)
|
7 |
|
8 |
+
|
9 |
|
10 |
+
passage_encoder = SentenceTransformer(
|
11 |
+
"facebook-dpr-ctx_encoder-multiset-base"
|
12 |
+
)
|
13 |
+
query_encoder = SentenceTransformer(
|
14 |
+
"facebook-dpr-question_encoder-multiset-base"
|
15 |
+
)
|
16 |
+
model = CrossEncoder("cross-encoder/ms-marco-MiniLM-L-12-v2")
|
17 |
+
embedder = SentenceTransformer("all-mpnet-base-v2")
|
18 |
+
|
19 |
+
|
20 |
+
def calc_preferred_dense(doc1, doc2, q1, q2, model_name="dpr"):
|
21 |
"""
|
22 |
Input:
|
23 |
doc1, doc2: strings containing the documents/passages
|
24 |
query1, query2: strings for queries that are only relevant to the corresponding doc (doc1 -> q1, doc2 -> q2)
|
25 |
model_name: string containing the type of model to run
|
|
|
26 |
|
27 |
Returns:
|
28 |
A dictionary containing each query (q1 or q2) and the score (P@1) for the pair
|
29 |
|
30 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
corpus = [doc1, doc2]
|
32 |
queries = [q1, q2]
|
33 |
results = {}
|
34 |
num_correct = 0
|
35 |
|
36 |
### Do Retrieval
|
37 |
+
if model_name == "dpr":
|
38 |
passage_embeddings = passage_encoder.encode(corpus)
|
39 |
|
40 |
query_encoder = SentenceTransformer(
|
|
|
53 |
num_correct += 1
|
54 |
model = (passage_encoder, query_encoder)
|
55 |
|
56 |
+
elif model_name == "cross_encoder":
|
57 |
for idx, query in enumerate(queries):
|
58 |
scores = model.predict([[query, doc1], [query, doc2]])
|
59 |
results[f"q{idx+1}"] = scores.tolist()
|
|
|
84 |
model = embedder
|
85 |
|
86 |
results["score"] = num_correct / 2
|
87 |
+
return results
|
88 |
|
89 |
|
90 |
+
gr.Interface(
|
91 |
+
calc_preferred_dense,
|
92 |
+
[ gr.Textbox(label="Sentence 1"), gr.Textbox(label="Sentence 2"), gr.Dropdown(["dpr", "cross-encoder", "dense"], value="cross-encoder")],
|
93 |
+
[ gr.components.Label(label="Similarity score") ],
|
94 |
+
title="Similarity score between 2 sentences",
|
95 |
+
description="In this demo do provide 2 sentences bellow. They can even be in distinct languages. Powered by S-BERT multilingual model : https://www.sbert.net.",
|
96 |
+
examples=[['The sentences are mapped such that sentences with similar meanings are close in vector space.', 'Les phrases sont mappées de manière à ce que les phrases ayant des significations similaires soient proches dans l\'espace vectoriel.'],
|
97 |
+
['You do not need to specify the input language.', 'You can use any language.']],
|
98 |
+
live=True,
|
99 |
+
allow_flagging="never"
|
100 |
+
).launch(debug=True, enable_queue=True)
|
101 |
|
|
|
|
requirements.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
sentence_transformers
|
3 |
+
numpy
|