Update app.py
Browse files
app.py
CHANGED
@@ -1,81 +1,81 @@
|
|
1 |
-
import streamlit as st
|
2 |
-
from multiprocessing import Process
|
3 |
-
|
4 |
-
from transformers import AutoConfig, AutoTokenizer, AutoModelForSequenceClassification
|
5 |
-
|
6 |
-
import torch
|
7 |
-
import pandas as pd
|
8 |
-
import json
|
9 |
-
import requests
|
10 |
-
|
11 |
-
import time
|
12 |
-
import os
|
13 |
-
|
14 |
-
model_name_or_directory = "MKaan/multilingual-cpv-sector-classifier"
|
15 |
-
tokenizer = AutoTokenizer.from_pretrained("bert-base-multilingual-cased")
|
16 |
-
|
17 |
-
config = AutoConfig.from_pretrained(model_name_or_directory)
|
18 |
-
model = AutoModelForSequenceClassification.from_pretrained(model_name_or_directory, config=config)
|
19 |
-
|
20 |
-
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
|
21 |
-
|
22 |
-
idx2cpv = pd.read_csv("idx2cpv.csv")
|
23 |
-
idx2cpv = dict(zip(idx2cpv.indexes, idx2cpv.sectors))
|
24 |
-
|
25 |
-
def get_result(input):
|
26 |
-
input_ids = tokenizer(input, return_tensors="pt").input_ids
|
27 |
-
output = model(input_ids)
|
28 |
-
pred = output.logits.argmax(dim=-1)
|
29 |
-
pred = pred.cpu().detach().numpy()[0]
|
30 |
-
return idx2cpv[pred]
|
31 |
-
|
32 |
-
if __name__ == "__main__":
|
33 |
-
st.title('Multilingual Sector Classifier 📄') #📊💼
|
34 |
-
st.subheader('Finds the correct sector for the given contract description')
|
35 |
-
st.markdown("Built by Mustafa Kaan Görgün, [Linkedin](https://www.linkedin.com/in/mustafa-kaan-görgün-a2461288/), [Model Card](https://huggingface.co/MKaan/multilingual-cpv-sector-classifier) ", unsafe_allow_html=True)
|
36 |
-
|
37 |
-
examples = pd.read_csv("examples.csv")
|
38 |
-
lang2example = dict(zip(examples.lang, examples.descr))
|
39 |
-
|
40 |
-
st.markdown(f'##### Try it now:')
|
41 |
-
|
42 |
-
#st.markdown(f'Choose a language in any of 22 languages')
|
43 |
-
input_lang = st.selectbox(
|
44 |
-
label="Choose a language from the list of 22 languages",
|
45 |
-
options=examples.lang,
|
46 |
-
index=5
|
47 |
-
)
|
48 |
-
|
49 |
-
input_text_1 = st.text_area(
|
50 |
-
label="Example description in choosen language",
|
51 |
-
value=lang2example[input_lang],
|
52 |
-
height=150,
|
53 |
-
max_chars=500
|
54 |
-
)
|
55 |
-
|
56 |
-
button1 = st.button('Run the example')
|
57 |
-
|
58 |
-
st.write("or")
|
59 |
-
|
60 |
-
#st.markdown('Write your own contract description in any of 104 languages that MBERT supports.')
|
61 |
-
input_text_2 = st.text_area(
|
62 |
-
label="Write your own contract description in any of 104 languages that MBERT supports.",
|
63 |
-
value="Your description comes here..",
|
64 |
-
height=100,
|
65 |
-
max_chars=500
|
66 |
-
)
|
67 |
-
|
68 |
-
button2 = st.button('Run your own')
|
69 |
-
|
70 |
-
st.markdown(f'##### Classified Sector: ')
|
71 |
-
if button1:
|
72 |
-
with st.spinner('In progress.......'):
|
73 |
-
sector_class = get_result(input_text_1)
|
74 |
-
#sector_class = input_text_1
|
75 |
-
st.success(sector_class)
|
76 |
-
|
77 |
-
if button2:
|
78 |
-
with st.spinner('In progress.......'):
|
79 |
-
sector_class = get_result(input_text_2)
|
80 |
-
#sector_class = input_text_2
|
81 |
st.success(sector_class)
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from multiprocessing import Process
|
3 |
+
|
4 |
+
from transformers import AutoConfig, AutoTokenizer, AutoModelForSequenceClassification
|
5 |
+
|
6 |
+
import torch
|
7 |
+
import pandas as pd
|
8 |
+
import json
|
9 |
+
import requests
|
10 |
+
|
11 |
+
import time
|
12 |
+
import os
|
13 |
+
|
14 |
+
model_name_or_directory = "MKaan/multilingual-cpv-sector-classifier"
|
15 |
+
tokenizer = AutoTokenizer.from_pretrained("bert-base-multilingual-cased")
|
16 |
+
|
17 |
+
config = AutoConfig.from_pretrained(model_name_or_directory)
|
18 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name_or_directory, config=config)
|
19 |
+
|
20 |
+
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
|
21 |
+
|
22 |
+
idx2cpv = pd.read_csv("idx2cpv.csv")
|
23 |
+
idx2cpv = dict(zip(idx2cpv.indexes, idx2cpv.sectors))
|
24 |
+
|
25 |
+
def get_result(input):
|
26 |
+
input_ids = tokenizer(input, return_tensors="pt").input_ids
|
27 |
+
output = model(input_ids)
|
28 |
+
pred = output.logits.argmax(dim=-1)
|
29 |
+
pred = pred.cpu().detach().numpy()[0]
|
30 |
+
return idx2cpv[pred]
|
31 |
+
|
32 |
+
if __name__ == "__main__":
|
33 |
+
st.title('Multilingual Sector Classifier 📄') #📊💼
|
34 |
+
st.subheader('Finds the correct sector for the given contract description')
|
35 |
+
st.markdown("Built by Mustafa Kaan Görgün, [Linkedin](https://www.linkedin.com/in/mustafa-kaan-görgün-a2461288/), [Model Card](https://huggingface.co/MKaan/multilingual-cpv-sector-classifier) ", unsafe_allow_html=True)
|
36 |
+
|
37 |
+
examples = pd.read_csv("examples.csv")
|
38 |
+
lang2example = dict(zip(examples.lang, examples.descr))
|
39 |
+
|
40 |
+
st.markdown(f'##### Try it now:')
|
41 |
+
|
42 |
+
#st.markdown(f'Choose a language in any of 22 languages')
|
43 |
+
input_lang = st.selectbox(
|
44 |
+
label="Choose a language from the list of 22 languages",
|
45 |
+
options=examples.lang,
|
46 |
+
index=5
|
47 |
+
)
|
48 |
+
|
49 |
+
input_text_1 = st.text_area(
|
50 |
+
label="Example description in choosen language",
|
51 |
+
value=lang2example[input_lang],
|
52 |
+
height=150,
|
53 |
+
max_chars=500
|
54 |
+
)
|
55 |
+
|
56 |
+
button1 = st.button('Run the example')
|
57 |
+
|
58 |
+
st.write("or")
|
59 |
+
|
60 |
+
#st.markdown('Write your own contract description in any of 104 languages that MBERT supports.')
|
61 |
+
input_text_2 = st.text_area(
|
62 |
+
label="Write your own contract description in any of 104 languages that MBERT supports.",
|
63 |
+
value="Your description comes here..",
|
64 |
+
height=100,
|
65 |
+
max_chars=500
|
66 |
+
)
|
67 |
+
|
68 |
+
button2 = st.button('Run your own')
|
69 |
+
|
70 |
+
st.markdown(f'##### Classified Sector: ')
|
71 |
+
if button1:
|
72 |
+
with st.spinner('In progress.......'):
|
73 |
+
sector_class = get_result(input_text_1)
|
74 |
+
#sector_class = input_text_1
|
75 |
+
st.success(sector_class)
|
76 |
+
|
77 |
+
if button2:
|
78 |
+
with st.spinner('In progress.......'):
|
79 |
+
sector_class = get_result(input_text_2)
|
80 |
+
#sector_class = input_text_2
|
81 |
st.success(sector_class)
|