Add application file
Browse files- app.py +56 -0
- requirements.txt +3 -0
app.py
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
from transformers import pipeline
|
3 |
+
from torch import Tensor
|
4 |
+
from transformers import AutoTokenizer, AutoModel
|
5 |
+
from torch.nn.functional import cosine_similarity
|
6 |
+
import gradio as gr
|
7 |
+
|
8 |
+
def average_pool(last_hidden_states: Tensor, attention_mask: Tensor) -> Tensor:
|
9 |
+
last_hidden = last_hidden_states.masked_fill(~attention_mask[..., None].bool(), 0.0)
|
10 |
+
return last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]
|
11 |
+
|
12 |
+
def get_similarity(sentence1, sentence2):
|
13 |
+
input_texts = [sentence1, sentence2]
|
14 |
+
# Tokenize and compute embeddings
|
15 |
+
batch_dict = tokenizer(input_texts, max_length=512, padding=True, truncation=True, return_tensors="pt")
|
16 |
+
outputs = model(**batch_dict)
|
17 |
+
embeddings = average_pool(outputs.last_hidden_state, batch_dict["attention_mask"])
|
18 |
+
similarity = cosine_similarity(embeddings[0].unsqueeze(0), embeddings[1].unsqueeze(0))
|
19 |
+
similarity = round(similarity.item(), 4)
|
20 |
+
return similarity
|
21 |
+
|
22 |
+
checkpoint = "intfloat/multilingual-e5-large"
|
23 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
24 |
+
model = AutoModel.from_pretrained(checkpoint)
|
25 |
+
|
26 |
+
demo = gr.Blocks(theme="freddyaboulton/dracula_revamped")
|
27 |
+
|
28 |
+
with demo:
|
29 |
+
gr.Markdown("# Sentence Similarity")
|
30 |
+
gr.Markdown("Task: Sentence Similarity (Multilingual)")
|
31 |
+
gr.Markdown("Model: https://huggingface.co/intfloat/multilingual-e5-large ")
|
32 |
+
|
33 |
+
p_txt1 = gr.Textbox(placeholder="Enter passage 1", label="Passage 1")
|
34 |
+
p_txt2 = gr.Textbox(placeholder="Enter passage 2", label="Passage 2")
|
35 |
+
submit = gr.Button("Submit")
|
36 |
+
o_txt = gr.Textbox(placeholder="Similarity score", lines=1,
|
37 |
+
interactive=False, label="Similarity score (0-10)")
|
38 |
+
|
39 |
+
gr.Examples(
|
40 |
+
[
|
41 |
+
["A big bus is running on the road in the city.", "There is a big bus running on the road."],
|
42 |
+
["A big bus is running on the road in the city.", "Two children in costumes are standing on the bed."],
|
43 |
+
["街中の道路を大きなバスが走っています。", "道路を大きなバスが走っています。"],
|
44 |
+
["街中の道路を大きなバスが走っています。", "ベッドの上で衣装を着た二人の子供が立っています。"],
|
45 |
+
["A big bus is running on the road in the city.", "道路を大きなバスが走っています。"]
|
46 |
+
],
|
47 |
+
inputs=[p_txt1, p_txt2]
|
48 |
+
)
|
49 |
+
|
50 |
+
submit.click(
|
51 |
+
get_similarity,
|
52 |
+
[p_txt1, p_txt2],
|
53 |
+
o_txt
|
54 |
+
)
|
55 |
+
|
56 |
+
demo.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
gradio==3.36.1
|
2 |
+
transformers==4.30.2
|
3 |
+
torch==2.0.1
|